Skip to content

diagnostics

zocrates.utils.diagnostics

Opt-in diagnostic snapshot builder and multi-layer privacy redaction engine.

Generates sanitized, GitHub-ready diagnostic reports strictly on user request. All disk and hardware inspection helpers are pure read-only operations safe for worker thread execution without mutating configuration files or databases.

ALLOWED_CONFIG_KEYS module-attribute 

ALLOWED_CONFIG_KEYS = frozenset({'llm_provider', 'llm_model', 'coordinator_thinking', 'subagent_thinking', 'embedding_mode', 'embedding_model', 'embedding_dimension', 'document_converter_id', 'docling_mode', 'datalab_merge_cross_page', 'citations_enabled', 'references_section_enabled', 'auto_render_mermaid_diagrams', 'enable_mcp_server', 'citation_verification_enabled', 'citation_verification_provider', 'citation_verification_model', 'citation_verification_threshold', 'adaptive_compaction_enabled', 'session_retention_days', 'session_max_count'})

BACKEND_PORT module-attribute 

BACKEND_PORT = _resolve_backend_port()

DEFAULT_CONFIG module-attribute 

DEFAULT_CONFIG = {'llm_provider': 'openrouter_llm', 'llm_model': DEFAULT_LLM_MODEL, 'llm_base_url': DEFAULT_LLM_BASE_URL, 'llm_temperature': DEFAULT_LLM_TEMPERATURE, 'llm_max_tokens': DEFAULT_LLM_MAX_TOKENS, 'llm_num_retries': DEFAULT_LLM_NUM_RETRIES, 'ollama_num_ctx': DEFAULT_OLLAMA_NUM_CTX, 'override_agent_models': copy.deepcopy(DEFAULT_OVERRIDE_AGENT_MODELS), 'embedding_provider': 'local', 'embedding_model': 'jinaai/jina-embeddings-v5-text-nano', 'embedding_dimension': 768, 'document_converter_mode': 'local', 'document_converter_id': 'markitdown', 'docling_mode': 'auto', 'datalab_non_pdf_fallback': 'markitdown', 'datalab_merge_cross_page': DEFAULT_DATALAB_MERGE_CROSS_PAGE, 'conversion_timeout_seconds': DEFAULT_CONVERSION_TIMEOUT_SECONDS, 'enable_vlm_captioning': DEFAULT_ENABLE_VLM_CAPTIONING, 'vlm_captioning_provider': DEFAULT_VLM_CAPTIONING_PROVIDER, 'vlm_captioning_model': DEFAULT_VLM_CAPTIONING_MODEL, 'vlm_captioning_backup_model': DEFAULT_VLM_CAPTIONING_BACKUP_MODEL, 'vlm_captioning_local_model': DEFAULT_VLM_CAPTIONING_LOCAL_MODEL, 'vlm_max_images_per_doc': DEFAULT_VLM_MAX_IMAGES_PER_DOC, 'vlm_timeout_seconds': DEFAULT_VLM_TIMEOUT_SECONDS, 'semantic_similarity_threshold': DEFAULT_SEMANTIC_SIMILARITY_THRESHOLD, 'reranker_model': DEFAULT_RERANKER_MODEL, 'search_harness': copy.deepcopy(DEFAULT_SEARCH_HARNESS), 'graph_relations': copy.deepcopy(DEFAULT_GRAPH_RELATIONS), 'citations_enabled': DEFAULT_CITATIONS_ENABLED, 'references_section_enabled': DEFAULT_REFERENCES_SECTION_ENABLED, 'considered_sources_disclosure': DEFAULT_CONSIDERED_SOURCES_DISCLOSURE, 'auto_render_mermaid_diagrams': DEFAULT_AUTO_RENDER_MERMAID_DIAGRAMS, 'enable_mcp_server': DEFAULT_ENABLE_MCP_SERVER, 'auto_spawn_backend': DEFAULT_AUTO_SPAWN_BACKEND, 'enable_predictive_ghost_input': DEFAULT_ENABLE_PREDICTIVE_GHOST_INPUT, 'predictive_prompt_model': DEFAULT_PREDICTIVE_PROMPT_MODEL, 'sentence_attribution_model': DEFAULT_SENTENCE_ATTRIBUTION_MODEL, 'scholar_polite_email': DEFAULT_SCHOLAR_POLITE_EMAIL, 'user_research_profile': DEFAULT_USER_RESEARCH_PROFILE, 'max_research_profile_chars': DEFAULT_MAX_RESEARCH_PROFILE_CHARS, 'citation_verification_enabled': False, 'citation_verification_provider': 'openrouter', 'citation_verification_model': DEFAULT_CITATION_VERIFICATION_MODEL, 'citation_verification_threshold': 0.6, 'citation_verification_local_url': 'http://127.0.0.1:8005/v1/systemone', 'citation_verification_local_port': 8005, 'citation_verification_auto_start_local': True, 'citation_verification_timeout_sec': 10.0, 'citation_verification_turn_budget_sec': 20.0, 'enable_adaptive_compaction': DEFAULT_ENABLE_ADAPTIVE_COMPACTION, 'adaptive_compaction_retention_turns': DEFAULT_ADAPTIVE_COMPACTION_RETENTION_TURNS, 'adaptive_compaction_min_turns': DEFAULT_ADAPTIVE_COMPACTION_MIN_TURNS, 'adaptive_compaction_token_floor': DEFAULT_ADAPTIVE_COMPACTION_TOKEN_FLOOR, 'adaptive_compaction_token_ceiling': DEFAULT_ADAPTIVE_COMPACTION_TOKEN_CEILING, 'adaptive_compaction_threshold': DEFAULT_ADAPTIVE_COMPACTION_THRESHOLD, 'adaptive_compaction_decision_model': DEFAULT_ADAPTIVE_COMPACTION_DECISION_MODEL, 'compaction_summarizer_model': DEFAULT_COMPACTION_SUMMARIZER_MODEL, 'max_saved_sessions': DEFAULT_MAX_SAVED_SESSIONS, 'session_retention_days': DEFAULT_SESSION_RETENTION_DAYS, 'session_db_auto_vacuum_on_startup': DEFAULT_SESSION_DB_AUTO_VACUUM_ON_STARTUP, 'prune_empty_sessions': DEFAULT_PRUNE_EMPTY_SESSIONS, 'backend_port': DEFAULT_BACKEND_PORT, 'bootstrap_handshake_timeout_ms': 300000, 'sync_auto_abort_failures': DEFAULT_SYNC_AUTO_ABORT_FAILURES}

DEFAULT_LOG_TAIL_LINES module-attribute 

DEFAULT_LOG_TAIL_LINES = 250

LOG_TAIL_SEEK_BYTES module-attribute 

LOG_TAIL_SEEK_BYTES = 256 * 1024

MAX_LOG_LINE_CHARS module-attribute 

MAX_LOG_LINE_CHARS = 500

MAX_LOG_TAIL_LINES module-attribute 

MAX_LOG_TAIL_LINES = 500

MAX_REPORT_BYTES module-attribute 

MAX_REPORT_BYTES = 50 * 1024

MIN_LOG_TAIL_LINES module-attribute 

MIN_LOG_TAIL_LINES = 50

OMITTED_LOGS_TEXT module-attribute 

OMITTED_LOGS_TEXT = '<Backend log tail omitted (enable via checkbox if needed for debugging)>'

OMITTED_PROMPTS_TEXT module-attribute 

OMITTED_PROMPTS_TEXT = '<User prompt and assistant message text omitted for privacy>'

OMITTED_TITLE_TEXT module-attribute 

OMITTED_TITLE_TEXT = '<Title omitted>'

TRUNCATION_MARKER module-attribute 

TRUNCATION_MARKER = '... [truncated to maintain 50 KB report ceiling]'

ZOCRATES_DATA_SUBDIR module-attribute 

ZOCRATES_DATA_SUBDIR = 'zocrates-data'

_BEARER_TOKEN_RE module-attribute 

_BEARER_TOKEN_RE = re.compile('(?i)(\\bBearer\\s+)(?!\\[REDACTED)([^\\s\\"\',;)}\\]]+)')

_CONFIG_CACHE module-attribute 

_CONFIG_CACHE = {}

_CONFIG_CACHE_FINGERPRINT module-attribute 

_CONFIG_CACHE_FINGERPRINT = {}

_CONFIG_CACHE_LOCK module-attribute 

_CONFIG_CACHE_LOCK = threading.Lock()

_EMAIL_RE module-attribute 

_EMAIL_RE = re.compile('\\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\\.[A-Za-z]{2,}\\b')

_GENERIC_SK_KEY_RE module-attribute 

_GENERIC_SK_KEY_RE = re.compile('\\bsk-[A-Za-z0-9_\\-]{16,}')

_LINUX_HOME_RE module-attribute 

_LINUX_HOME_RE = re.compile('(?<![A-Za-z0-9_\\-])(/home/)([^/\\s\\"\':;<>|\\\\]+)')

_LINUX_VAR_HOME_RE module-attribute 

_LINUX_VAR_HOME_RE = re.compile('(?<![A-Za-z0-9_\\-])(/var/home/)([^/\\s\\"\':;<>|\\\\]+)')

_MACOS_USERS_RE module-attribute 

_MACOS_USERS_RE = re.compile('(?<![A-Za-z0-9_\\-])(/Users/)([^/\\s\\"\':;<>|\\\\]+)')

_MAILTO_RE module-attribute 

_MAILTO_RE = re.compile('(?i)\\bmailto:[^\\s\\"\',;>)\\]]+')

_OPENROUTER_KEY_RE module-attribute 

_OPENROUTER_KEY_RE = re.compile('sk-or-v1-[A-Za-z0-9_\\-]{8,}')

_QUERY_TOKEN_RE module-attribute 

_QUERY_TOKEN_RE = re.compile('(?i)([?&]?(?:token|api_key|apikey|access_token|auth_token)=)(?!\\[REDACTED)([^\\s&\\"\',;)}\\]]+)')

_URL_USERINFO_RE module-attribute 

_URL_USERINFO_RE = re.compile('(https?://)([^/\\s:@]+:[^/\\s@]+)@')

_WIN_BACKSLASH_USERS_RE module-attribute 

_WIN_BACKSLASH_USERS_RE = re.compile('(?<![A-Za-z0-9_\\-])[A-Za-z]:\\\\Users\\\\[^\\\\\\s\\"\':;<>|/]+')

_WIN_ESCAPED_UNC_RE module-attribute 

_WIN_ESCAPED_UNC_RE = re.compile('(?<![\\\\/\\w:])\\\\\\\\\\\\\\\\[A-Za-z0-9._\\-]+\\\\\\\\[A-Za-z0-9._$\\-]+')

_WIN_ESCAPED_USERS_RE module-attribute 

_WIN_ESCAPED_USERS_RE = re.compile('(?<![A-Za-z0-9_\\-])[A-Za-z]:\\\\\\\\Users\\\\\\\\[^\\\\\\s\\"\':;<>|/]+')

_WIN_SLASH_USERS_RE module-attribute 

_WIN_SLASH_USERS_RE = re.compile('(?<![A-Za-z0-9_\\-])[A-Za-z]:/Users/[^/\\s\\"\':;<>|\\\\]+')

_WIN_UNC_RE module-attribute 

_WIN_UNC_RE = re.compile('(?<![\\\\/\\w:])\\\\\\\\[A-Za-z0-9._\\-]+\\\\[A-Za-z0-9._$\\-]+')

_X_API_KEY_HEADER_RE module-attribute 

_X_API_KEY_HEADER_RE = re.compile('(?i)(\\bX-Api-Key\\s*[:=]\\s*)(?!\\[REDACTED)([^\\s\\"\',;)}\\]]+)')

_ZOCRATES_AUTH_HEADER_RE module-attribute 

_ZOCRATES_AUTH_HEADER_RE = re.compile('(?i)(\\bX-Zocrates-Auth\\s*[:=]\\s*)(?!\\[REDACTED)([^\\s\\"\',;)}\\]]+)')

__version__ module-attribute 

__version__ = version('zocrates')

logger module-attribute 

logger = logging.getLogger(__name__)

_collect_exact_secrets 

_collect_exact_secrets(resolved_dir)

Collect active runtime secrets as (secret_value, redaction_placeholder) pairs.

Source code in zocrates/utils/diagnostics.py 
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
def _collect_exact_secrets(resolved_dir: Path | None) -> list[tuple[str, str]]:
    """Collect active runtime secrets as (secret_value, redaction_placeholder) pairs."""
    candidates: list[tuple[str, str]] = []

    def _add(val: str | None, placeholder: str) -> None:
        if not val or not isinstance(val, str):
            return
        cleaned = val.strip()
        if len(cleaned) >= 4:
            candidates.append((cleaned, placeholder))

    # 1. Backend auth token (purely read-only: never create backend.auth or mutate _CACHED_AUTH_TOKEN).
    with contextlib.suppress(Exception):
        env_tok = os.getenv("ZOCRATES_AUTH_TOKEN", "").strip()
        if env_tok:
            _add(env_tok, "[REDACTED_AUTH_TOKEN]")
        cached_tok = getattr(_zocrates_cfg, "_CACHED_AUTH_TOKEN", None)
        if isinstance(cached_tok, str) and cached_tok.strip():
            _add(cached_tok, "[REDACTED_AUTH_TOKEN]")
        if resolved_dir is not None:
            sidecar_path = resolved_dir / ZOCRATES_DATA_SUBDIR / "backend.auth"
            if sidecar_path.is_file():
                _add(
                    sidecar_path.read_text(encoding="utf-8").strip(),
                    "[REDACTED_AUTH_TOKEN]",
                )

    # 2. Live API keys from environment/memory.
    with contextlib.suppress(Exception):
        _add(get_secure_api_key(), "[REDACTED_API_KEY]")
    with contextlib.suppress(Exception):
        _add(
            get_secure_api_key("openai_api_key", ("OPENAI_API_KEY",)),
            "[REDACTED_API_KEY]",
        )
    with contextlib.suppress(Exception):
        _add(get_datalab_api_key(resolved_dir), "[REDACTED_API_KEY]")
    with contextlib.suppress(Exception):
        _add(get_openalex_api_key(resolved_dir), "[REDACTED_API_KEY]")

    # 3. Configured polite-pool email if present in config.
    with contextlib.suppress(Exception):
        raw_cfg = _read_config_readonly(resolved_dir)
        polite_email = raw_cfg.get("scholar_polite_email")
        if isinstance(polite_email, str) and polite_email.strip():
            _add(polite_email, "[REDACTED_EMAIL]")

    # Sort longest-first so shorter substrings never partially clobber longer keys.
    dedup: dict[str, str] = {}
    for secret, placeholder in sorted(candidates, key=lambda item: len(item[0]), reverse=True):
        if secret not in dedup:
            dedup[secret] = placeholder
    return list(dedup.items())

_collect_path_variants 

_collect_path_variants(resolved_dir)

Pre-collect path variants for resolved_dir and Path.home().

Source code in zocrates/utils/diagnostics.py 
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
def _collect_path_variants(resolved_dir: Path | None) -> list[tuple[str, str]]:
    """Pre-collect path variants for resolved_dir and Path.home()."""
    variants: list[tuple[str, str]] = []
    if resolved_dir is not None:
        paths_to_check = [resolved_dir]
        with contextlib.suppress(OSError):
            r = resolved_dir.resolve()
            if r != resolved_dir:
                paths_to_check.append(r)
        for p in paths_to_check:
            native = str(p).rstrip("/\\")
            posix = p.as_posix().rstrip("/")
            for candidate in (native.replace("\\", "\\\\"), native, posix):
                if candidate and len(candidate) > 2 and candidate not in ("/", "\\", "."):
                    if not any(candidate == v[0] for v in variants):
                        variants.append((candidate, "<ZOTERO_DATA_DIR>"))

    with contextlib.suppress(Exception):
        home_path = Path.home()
        paths_to_check = [home_path]
        with contextlib.suppress(OSError):
            r = home_path.resolve()
            if r != home_path:
                paths_to_check.append(r)
        for p in paths_to_check:
            native = str(p).rstrip("/\\")
            posix = p.as_posix().rstrip("/")
            for candidate in (native.replace("\\", "\\\\"), native, posix):
                if candidate and len(candidate) > 2 and candidate not in ("/", "\\", "."):
                    if not any(candidate == v[0] for v in variants):
                        variants.append((candidate, "<HOME>"))

    variants.sort(key=lambda item: len(item[0]), reverse=True)
    return variants

_format_context_chips 

_format_context_chips(raw_items, *, include_item_titles, zotero_data_dir=None)

Format attached context chips while honoring include_item_titles.

Source code in zocrates/utils/diagnostics.py 
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
def _format_context_chips(
    raw_items: list[Any] | None,
    *,
    include_item_titles: bool,
    zotero_data_dir: Path | str | None = None,
) -> list[str]:
    """Format attached context chips while honoring ``include_item_titles``."""
    if not raw_items or not isinstance(raw_items, list):
        return []

    formatted: list[str] = []
    for item in raw_items[:25]:
        if not isinstance(item, dict):
            continue
        raw_key = (
            item.get("key")
            or item.get("item_key")
            or item.get("itemKey")
            or "unknown"
        )
        safe_key = sanitize_text(str(raw_key), zotero_data_dir=zotero_data_dir)[:32]
        if include_item_titles:
            raw_title = str(item.get("title") or "Untitled")
            clean_title = sanitize_text(raw_title, zotero_data_dir=zotero_data_dir)
            formatted.append(f"[Key: {safe_key}] {clean_title}")
        else:
            formatted.append(f"[Key: {safe_key}] {OMITTED_TITLE_TEXT}")
    return formatted

_is_compaction_event 

_is_compaction_event(e)

Return True if event is an ADK compaction event, ignoring unconfigured test mocks.

Source code in zocrates/utils/session_formatting.py 
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
def _is_compaction_event(e: Any) -> bool:
    """Return True if event is an ADK compaction event, ignoring unconfigured test mocks."""
    actions = getattr(e, "actions", None)
    if actions is None:
        return False
    if isinstance(actions, MagicMock) and "compaction" not in actions.__dict__:
        return False
    compaction = getattr(actions, "compaction", None)
    if compaction is None:
        return False
    if (
        isinstance(compaction, MagicMock)
        and "compacted_content" not in compaction.__dict__
        and "compaction" not in getattr(actions, "__dict__", {})
    ):
        return False
    return bool(compaction)

_read_config_readonly 

_read_config_readonly(zotero_data_dir=None)

Read Zocrates configuration without acquiring FileLock or writing backfills to disk.

Source code in zocrates/utils/diagnostics.py 
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
def _read_config_readonly(zotero_data_dir: Path | str | None = None) -> dict[str, Any]:
    """Read Zocrates configuration without acquiring FileLock or writing backfills to disk."""
    cfg = copy.deepcopy(DEFAULT_CONFIG)
    resolved_dir = _resolve_zotero_dir_readonly(zotero_data_dir)
    if resolved_dir is None:
        return cfg

    config_path = resolved_dir / ZOCRATES_DATA_SUBDIR / "zocrates_config.json"
    if not config_path.is_file():
        return cfg

    try:
        resolved_path = config_path.resolve()
        stat_res = config_path.stat()
        fingerprint = (stat_res.st_mtime_ns, stat_res.st_size)
        with _CONFIG_CACHE_LOCK:
            if (
                resolved_path in _CONFIG_CACHE
                and _CONFIG_CACHE_FINGERPRINT.get(resolved_path) == fingerprint
            ):
                cached = copy.deepcopy(_CONFIG_CACHE[resolved_path])
                cfg.update(cached)
                return cfg
    except OSError:
        return cfg

    try:
        with open(config_path, encoding="utf-8") as f:
            raw = json.load(f)
        if isinstance(raw, dict):
            cfg.update(raw)
    except (OSError, json.JSONDecodeError):
        pass
    return cfg

_read_tail_lines_binary 

_read_tail_lines_binary(file_path, max_lines, max_bytes=LOG_TAIL_SEEK_BYTES)

Read up to max_lines from the end of file_path via binary EOF seek.

Source code in zocrates/utils/diagnostics.py 
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
def _read_tail_lines_binary(
    file_path: Path,
    max_lines: int,
    max_bytes: int = LOG_TAIL_SEEK_BYTES,
) -> list[str]:
    """Read up to ``max_lines`` from the end of ``file_path`` via binary EOF seek."""
    try:
        if not file_path.is_file():
            return []
        with open(file_path, "rb") as f:
            f.seek(0, os.SEEK_END)
            file_size = f.tell()
            if file_size <= 0:
                return []
            start_pos = max(0, file_size - max_bytes)
            f.seek(start_pos, os.SEEK_SET)
            chunk = f.read(file_size - start_pos)
    except (OSError, ValueError):
        return []

    decoded = chunk.decode("utf-8", errors="replace")
    lines = decoded.splitlines()
    # Drop potentially mid-line first fragment when seeking into the middle of a large file.
    if start_pos > 0 and len(lines) > 1:
        lines = lines[1:]
    return lines[-max_lines:] if max_lines > 0 else []

_render_markdown_from_parts 

_render_markdown_from_parts(data, *, log_lines, trace_entries, max_prompt_chars=None, logs_truncated=False, trace_truncated=False)

Render GitHub-flavored Markdown report with collapsible <details> sections.

Source code in zocrates/utils/diagnostics.py 
 979
 980
 981
 982
 983
 984
 985
 986
 987
 988
 989
 990
 991
 992
 993
 994
 995
 996
 997
 998
 999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
def _render_markdown_from_parts(
    data: dict[str, Any],
    *,
    log_lines: list[str] | None,
    trace_entries: list[dict[str, Any]],
    max_prompt_chars: int | None = None,
    logs_truncated: bool = False,
    trace_truncated: bool = False,
) -> str:
    """Render GitHub-flavored Markdown report with collapsible ``<details>`` sections."""
    env = data.get("environment") or {}
    fe = data.get("frontend") or {}
    cfg = data.get("config") or {}
    key_status = data.get("api_key_status") or {}
    turn = data.get("turn_context") or {}
    logs_info = data.get("logs") or {}

    hw = env.get("hardware") or {}
    docling = env.get("docling_assessment") or {}

    err_msg = turn.get("error_message") or "No specific error message attached (manual report)."
    if max_prompt_chars is not None and len(err_msg) > max_prompt_chars:
        err_msg = err_msg[:max_prompt_chars] + f"\n{TRUNCATION_MARKER}"

    latency_val = fe.get("heartbeat_latency_ms")
    latency_str = f"{latency_val} ms" if latency_val is not None else "N/A"

    vram_val = hw.get("vram_gb")
    vram_str = f"{vram_val} GB" if vram_val is not None else "N/A"
    cc_val = hw.get("compute_capability")
    cc_str = str(cc_val) if cc_val is not None else "N/A"

    lines: list[str] = [
        "## Zocrates Diagnostic & Error Report",
        "",
        f"- **Generated**: `{data.get('generated_at', 'Unknown')}`",
        f"- **Zocrates Backend Version**: `{env.get('backend_version', 'Unknown')}`",
        f"- **Zocrates Extension Version**: `{fe.get('extension_version', '0.1.0')}`",
        f"- **Zotero Version**: `{fe.get('zotero_version', 'Unknown')}` (`{fe.get('zotero_platform', 'Unknown')}`)",
        "",
        "### Error Summary",
        "```text",
        str(err_msg),
        "```",
        "",
        "### Environment & Hardware",
        f"- **OS / Architecture**: `{env.get('os_platform', 'Unknown')} {env.get('os_release', '')}` (`{env.get('architecture', 'Unknown')}`)",
        f"- **Python Runtime**: `{env.get('python_version', 'Unknown')}`",
        f"- **Window Mode**: `{fe.get('window_mode', 'sidebar')}`",
        f"- **Backend Connection**: `{fe.get('backend_connection', 'Connected')}` (Port: `{env.get('backend_port', 'Unknown')}`, PID: `{env.get('backend_pid', 'Unknown')}`, Latency: `{latency_str}`)",
        f"- **System RAM**: `{hw.get('ram_gb', 'Unknown')} GB`",
        f"- **GPU**: `{hw.get('gpu_vendor', 'cpu')}` — `{hw.get('gpu_name', 'Unknown')}` (VRAM: `{vram_str}`, Compute Capability: `{cc_str}`)",
        f"- **Docling Assessment**: `{docling.get('recommended_mode', 'N/A')}` — {docling.get('reason', 'N/A')}",
        "",
        "<details>",
        "<summary><b>Active Configuration (Sanitized Allowlist)</b></summary>",
        "",
        "```json",
        json.dumps(
            {"config": cfg, "api_key_status": key_status},
            indent=2,
            sort_keys=True,
        ),
        "```",
        "",
        "</details>",
        "",
        "<details>",
        "<summary><b>Turn & Execution Context</b></summary>",
        "",
        f"- **Session ID**: `{turn.get('session_id') or 'None'}`",
        f"- **Invocation ID**: `{turn.get('invocation_id') or 'None'}`",
    ]

    chips = turn.get("context_chips") or []
    if chips:
        lines.append("- **Context Chips**:")
        for chip in chips:
            lines.append(f"  - `{chip}`")
    else:
        lines.append("- **Context Chips**: `None`")

    prompts = turn.get("prompts") or {}
    if prompts.get("included"):
        u_prompt = str(prompts.get("user_prompt") or "<Empty>")
        a_resp = str(prompts.get("assistant_response") or "<Empty>")
        if max_prompt_chars is not None:
            if len(u_prompt) > max_prompt_chars:
                u_prompt = u_prompt[:max_prompt_chars] + f"\n{TRUNCATION_MARKER}"
            if len(a_resp) > max_prompt_chars:
                a_resp = a_resp[:max_prompt_chars] + f"\n{TRUNCATION_MARKER}"
        lines.extend(
            [
                "",
                "**User Prompt**:",
                "```text",
                u_prompt,
                "```",
                "",
                "**Assistant / Error Output**:",
                "```text",
                a_resp,
                "```",
            ]
        )
    else:
        lines.append(
            f"- **Prompt & Response**: `{prompts.get('placeholder', OMITTED_PROMPTS_TEXT)}`"
        )

    lines.append("")
    if trace_entries or trace_truncated:
        lines.append("**Shallow Tool Execution Trace**:")
        for step in trace_entries:
            keys_str = ", ".join(step.get("arg_keys") or [])
            err_suffix = f" — error: `{step['error']}`" if step.get("error") else ""
            lines.append(
                f"- `{step.get('tool', 'tool')}` (author: `{step.get('author', 'Agent')}`, "
                f"arg_keys: `[{keys_str}]`) -> `{step.get('status', 'ok')}`{err_suffix}"
            )
        if trace_truncated:
            lines.append(f"- `{TRUNCATION_MARKER}`")
    else:
        lines.append("- **Shallow Tool Execution Trace**: `None recorded`")

    lines.extend(["", "</details>", ""])

    notifs = fe.get("notifications") or []
    lines.extend(
        [
            "<details>",
            "<summary><b>Frontend Notifications (Last 5)</b></summary>",
            "",
        ]
    )
    if notifs:
        for n in notifs[:5]:
            ts_prefix = f"[{n['timestamp']}] " if n.get("timestamp") else ""
            lines.append(f"- `{n.get('type', 'info')}`: {ts_prefix}{n.get('text', '')}")
    else:
        lines.append("_No recent notifications._")
    lines.extend(["", "</details>", ""])

    lines.extend(
        [
            "<details>",
            "<summary><b>Backend Log Tail</b></summary>",
            "",
            "```text",
        ]
    )
    if logs_info.get("included"):
        if logs_truncated:
            lines.append(TRUNCATION_MARKER)
        if log_lines:
            lines.extend(log_lines)
        elif not logs_truncated:
            lines.append("<No log entries found in zocrates.log>")
    else:
        lines.append(str(logs_info.get("placeholder") or OMITTED_LOGS_TEXT))

    lines.extend(["```", "", "</details>", ""])
    return "\n".join(lines)

_replace_path_variants 

_replace_path_variants(text, target_path, placeholder)

Replace native, POSIX, and JSON-escaped variants of target_path in text.

Source code in zocrates/utils/diagnostics.py 
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
def _replace_path_variants(text: str, target_path: Path, placeholder: str) -> str:
    """Replace native, POSIX, and JSON-escaped variants of ``target_path`` in ``text``."""
    variants: list[str] = []
    paths_to_check = [target_path]
    with contextlib.suppress(OSError):
        resolved = target_path.resolve()
        if resolved != target_path:
            paths_to_check.append(resolved)

    for p in paths_to_check:
        native = str(p).rstrip("/\\")
        posix = p.as_posix().rstrip("/")
        for candidate in (
            native.replace("\\", "\\\\"),
            native,
            posix,
        ):
            if candidate and candidate not in ("/", "\\", ".") and len(candidate) > 2:
                if candidate not in variants:
                    variants.append(candidate)

    variants.sort(key=len, reverse=True)
    for variant in variants:
        if variant in text:
            text = text.replace(variant, placeholder)
    return text

_resolve_zotero_dir_readonly 

_resolve_zotero_dir_readonly(zotero_data_dir=None)

Resolve Zotero data directory without creating directories on disk.

Source code in zocrates/utils/diagnostics.py 
135
136
137
138
139
140
141
142
143
def _resolve_zotero_dir_readonly(zotero_data_dir: Path | str | None = None) -> Path | None:
    """Resolve Zotero data directory without creating directories on disk."""
    if zotero_data_dir is not None:
        return Path(zotero_data_dir)
    with contextlib.suppress(Exception):
        resolved = resolve_zotero_data_dir(None)
        if resolved is not None:
            return Path(resolved)
    return None

_sanitize_frontend_metadata 

_sanitize_frontend_metadata(frontend_metadata, *, include_prompts, include_item_titles, zotero_data_dir=None)

Sanitize untrusted client-provided diagnostic metadata and notifications.

Source code in zocrates/utils/diagnostics.py 
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
def _sanitize_frontend_metadata(
    frontend_metadata: dict[str, Any] | None,
    *,
    include_prompts: bool,
    include_item_titles: bool,
    zotero_data_dir: Path | str | None = None,
) -> dict[str, Any]:
    """Sanitize untrusted client-provided diagnostic metadata and notifications."""
    if not isinstance(frontend_metadata, dict):
        return {
            "zotero_version": "Unknown",
            "zotero_platform": "Unknown",
            "extension_version": "0.1.0",
            "window_mode": "unknown",
            "backend_connection": "Connected",
            "heartbeat_latency_ms": None,
            "notifications": [],
        }

    def _clean_short(key: str, default: str = "Unknown", max_len: int = 120) -> str:
        val = frontend_metadata.get(key)
        if val is None or val == "":
            return default
        return sanitize_text(str(val), zotero_data_dir=zotero_data_dir)[:max_len]

    raw_latency = frontend_metadata.get("heartbeat_latency_ms")
    latency_ms: float | None = None
    if isinstance(raw_latency, (int, float)) and not isinstance(raw_latency, bool) and raw_latency >= 0:
        latency_ms = round(float(raw_latency), 1)

    # Cap notifications at last 5 and scrub titles/queries when toggles are false.
    raw_notifs = frontend_metadata.get("notifications")
    clean_notifs: list[dict[str, str]] = []
    if isinstance(raw_notifs, list):
        for item in raw_notifs[:5]:
            if not isinstance(item, dict):
                continue
            n_type = sanitize_text(str(item.get("type") or "info"), zotero_data_dir=zotero_data_dir)[:24]
            n_text = sanitize_text(str(item.get("text") or ""), zotero_data_dir=zotero_data_dir)
            if not include_item_titles:
                # Scrub paper titles embedded in auto/batch converter notifications.
                n_text = re.sub(
                    r"((?:âš¡\s*)?(?:Re-converting|Converting)(?:\s+new)?(?:\s+[^:]*)?:\s*)(.+?)(\s+with\s+)",
                    rf"\1{OMITTED_TITLE_TEXT}\3",
                    n_text,
                )
                n_text = re.sub(
                    r"(Indexing failed:\s*)(.+?)(\s+\()",
                    rf"\1{OMITTED_TITLE_TEXT}\3",
                    n_text,
                )
                # Scrub double-quoted note/paper titles (e.g. Saved "Note" attached directly under "Paper").
                n_text = re.sub(
                    r'"([^"\n\r]{2,})"',
                    rf'"{OMITTED_TITLE_TEXT}"',
                    n_text,
                )
            if not include_prompts and len(n_text) > 240:
                n_text = n_text[:237] + "..."
            else:
                n_text = n_text[:400]
            n_ts = sanitize_text(str(item.get("timestamp") or ""), zotero_data_dir=zotero_data_dir)[:40]
            clean_notifs.append({"type": n_type, "text": n_text, "timestamp": n_ts})

    return {
        "zotero_version": _clean_short("zotero_version", "Unknown"),
        "zotero_platform": _clean_short("zotero_platform", "Unknown"),
        "extension_version": _clean_short("extension_version", "0.1.0"),
        "window_mode": _clean_short("window_mode", "sidebar"),
        "backend_connection": _clean_short("backend_connection", "Connected"),
        "heartbeat_latency_ms": latency_ms,
        "notifications": clean_notifs,
    }

build_diagnostic_report_bundle 

build_diagnostic_report_bundle(*, zotero_data_dir=None, session=None, session_id=None, invocation_id=None, error_message=None, include_prompts=False, include_item_titles=False, include_logs=False, log_tail_lines=DEFAULT_LOG_TAIL_LINES, frontend_metadata=None)

Build both the structured diagnostic snapshot and the 50 KB-capped Markdown report.

Source code in zocrates/utils/diagnostics.py 
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
def build_diagnostic_report_bundle(
    *,
    zotero_data_dir: Path | str | None = None,
    session: Any | None = None,
    session_id: str | None = None,
    invocation_id: str | None = None,
    error_message: str | None = None,
    include_prompts: bool = False,
    include_item_titles: bool = False,
    include_logs: bool = False,
    log_tail_lines: int = DEFAULT_LOG_TAIL_LINES,
    frontend_metadata: dict[str, Any] | None = None,
) -> dict[str, Any]:
    """Build both the structured diagnostic snapshot and the 50 KB-capped Markdown report."""
    snapshot = build_diagnostic_snapshot(
        zotero_data_dir=zotero_data_dir,
        session=session,
        session_id=session_id,
        invocation_id=invocation_id,
        error_message=error_message,
        include_prompts=include_prompts,
        include_item_titles=include_item_titles,
        include_logs=include_logs,
        log_tail_lines=log_tail_lines,
        frontend_metadata=frontend_metadata,
    )
    markdown = format_markdown_report(snapshot)
    return {
        "status": "success",
        "markdown": markdown,
        "report": snapshot,
    }

build_diagnostic_snapshot 

build_diagnostic_snapshot(*, zotero_data_dir=None, session=None, session_id=None, invocation_id=None, error_message=None, include_prompts=False, include_item_titles=False, include_logs=False, log_tail_lines=DEFAULT_LOG_TAIL_LINES, frontend_metadata=None)

Assemble a sanitized diagnostic snapshot dictionary on a worker thread.

Source code in zocrates/utils/diagnostics.py 
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
def build_diagnostic_snapshot(
    *,
    zotero_data_dir: Path | str | None = None,
    session: Any | None = None,
    session_id: str | None = None,
    invocation_id: str | None = None,
    error_message: str | None = None,
    include_prompts: bool = False,
    include_item_titles: bool = False,
    include_logs: bool = False,
    log_tail_lines: int = DEFAULT_LOG_TAIL_LINES,
    frontend_metadata: dict[str, Any] | None = None,
) -> dict[str, Any]:
    """Assemble a sanitized diagnostic snapshot dictionary on a worker thread."""
    resolved_dir = _resolve_zotero_dir_readonly(zotero_data_dir)
    generated_at = datetime.datetime.now(datetime.timezone.utc).strftime(
        "%Y-%m-%dT%H:%M:%SZ"
    )

    # 1. Hardware & OS profile (cached, no cold subprocess probes if already cached).
    hw_dict: dict[str, Any] = {}
    docling_assessment: dict[str, Any] = {}
    with contextlib.suppress(Exception):
        hw_dict = get_system_hardware_info(force_refresh=False).to_dict()
    with contextlib.suppress(Exception):
        docling_assessment = get_docling_hardware_assessment()
        if isinstance(docling_assessment.get("reason"), str):
            docling_assessment = dict(docling_assessment)
            docling_assessment["reason"] = sanitize_text(
                docling_assessment["reason"], zotero_data_dir=resolved_dir
            )

    environment = {
        "os_platform": sanitize_text(platform.system(), zotero_data_dir=resolved_dir),
        "os_release": sanitize_text(platform.release(), zotero_data_dir=resolved_dir),
        "architecture": sanitize_text(platform.machine(), zotero_data_dir=resolved_dir),
        "python_version": sanitize_text(
            sys.version.split()[0], zotero_data_dir=resolved_dir
        ),
        "backend_version": sanitize_text(__version__, zotero_data_dir=resolved_dir),
        "backend_port": BACKEND_PORT,
        "backend_pid": os.getpid(),
        "hardware": hw_dict,
        "docling_assessment": docling_assessment,
    }

    # 2. Strict configuration allowlist + boolean API key presence flags.
    config_allowlist = extract_sanitized_config(resolved_dir)
    api_key_status = {
        "has_openrouter_key": bool(get_secure_api_key()),
        "has_datalab_key": bool(get_datalab_api_key(resolved_dir)),
        "has_openalex_key": bool(get_openalex_api_key(resolved_dir)),
    }

    # 3. Frontend context.
    client_info = _sanitize_frontend_metadata(
        frontend_metadata,
        include_prompts=include_prompts,
        include_item_titles=include_item_titles,
        zotero_data_dir=resolved_dir,
    )

    # 4. Shallow turn context & trace.
    fallback_items = (
        frontend_metadata.get("active_items")
        if isinstance(frontend_metadata, dict)
        else None
    )
    turn_context = extract_shallow_turn_context(
        session,
        invocation_id=invocation_id,
        error_message=error_message,
        include_prompts=include_prompts,
        include_item_titles=include_item_titles,
        fallback_active_items=fallback_items,
        zotero_data_dir=resolved_dir,
    )
    if not turn_context.get("session_id") and session_id:
        turn_context["session_id"] = sanitize_text(
            session_id, zotero_data_dir=resolved_dir
        )

    # Allow client-supplied prompt fallback when session had not persisted yet AND include_prompts=True.
    if (
        include_prompts
        and isinstance(frontend_metadata, dict)
        and isinstance(turn_context.get("prompts"), dict)
        and not turn_context["prompts"].get("user_prompt")
        and frontend_metadata.get("last_user_prompt")
    ):
        turn_context["prompts"]["user_prompt"] = sanitize_text(
            str(frontend_metadata["last_user_prompt"]),
            zotero_data_dir=resolved_dir,
        )

    # 5. Backend log tail (strictly opt-in).
    if include_logs:
        logs_section: dict[str, Any] = {
            "included": True,
            "lines": get_sanitized_log_tail(resolved_dir, max_lines=log_tail_lines),
        }
    else:
        logs_section = {
            "included": False,
            "placeholder": OMITTED_LOGS_TEXT,
            "lines": [],
        }

    return {
        "generated_at": generated_at,
        "toggles": {
            "include_prompts": bool(include_prompts),
            "include_item_titles": bool(include_item_titles),
            "include_logs": bool(include_logs),
        },
        "environment": environment,
        "frontend": client_info,
        "config": config_allowlist,
        "api_key_status": api_key_status,
        "turn_context": turn_context,
        "logs": logs_section,
    }

extract_event_parts 

extract_event_parts(evt)

Extract text, thoughts, function calls, and responses from an event.

Source code in zocrates/utils/session_formatting.py 
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
def extract_event_parts(evt: Any) -> tuple[str, list[str], list[dict[str, Any]], list[dict[str, Any]]]:
    """Extract text, thoughts, function calls, and responses from an event."""
    content = getattr(evt, "content", None)
    parts = getattr(content, "parts", None) if content else None
    texts: list[str] = []
    thoughts: list[str] = []
    calls: list[dict[str, Any]] = []
    responses: list[dict[str, Any]] = []

    if isinstance(content, str) and content.strip():
        texts.append(content.strip())
    elif parts:
        current_thought_chunk: list[str] = []
        for p in parts:
            thought_val = getattr(p, "thought", None)
            if bool(thought_val):
                if isinstance(thought_val, str) and thought_val:
                    current_thought_chunk.append(thought_val)
                elif hasattr(p, "text") and p.text and not isinstance(p.text, MagicMock):
                    current_thought_chunk.append(str(p.text))
            else:
                if current_thought_chunk:
                    joined_th = "".join(current_thought_chunk).strip()
                    if joined_th:
                        thoughts.append(joined_th)
                    current_thought_chunk = []

                fc = getattr(p, "function_call", None)
                if fc is not None and not isinstance(fc, MagicMock) and getattr(fc, "name", None):
                    calls.append({
                        "name": fc.name,
                        "args": getattr(fc, "args", {}) or {},
                        "id": getattr(fc, "id", None),
                    })
                fr = getattr(p, "function_response", None)
                if fr is not None and not isinstance(fr, MagicMock) and getattr(fr, "name", None):
                    responses.append({
                        "name": fr.name,
                        "response": getattr(fr, "response", None),
                        "id": getattr(fr, "id", None),
                    })
                if hasattr(p, "text") and p.text and not isinstance(p.text, MagicMock) and str(p.text).strip():
                    texts.append(str(p.text).strip())

        if current_thought_chunk:
            joined_th = "".join(current_thought_chunk).strip()
            if joined_th:
                thoughts.append(joined_th)

    return "\n".join(texts), thoughts, calls, responses

extract_sanitized_config 

extract_sanitized_config(zotero_data_dir=None, raw_cfg=None)

Serialize strictly allowlisted technical configuration keys and boolean key flags.

Source code in zocrates/utils/diagnostics.py 
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
def extract_sanitized_config(
    zotero_data_dir: Path | str | None = None,
    raw_cfg: dict[str, Any] | None = None,
) -> dict[str, Any]:
    """Serialize strictly allowlisted technical configuration keys and boolean key flags."""
    resolved_dir = _resolve_zotero_dir_readonly(zotero_data_dir)
    cfg = copy.deepcopy(raw_cfg) if raw_cfg is not None else _read_config_readonly(resolved_dir)

    override_models = (
        cfg.get("override_agent_models")
        if isinstance(cfg.get("override_agent_models"), dict)
        else {}
    )
    coordinator_spec = (
        override_models.get("coordinator")
        if isinstance(override_models.get("coordinator"), dict)
        else {}
    )
    subagent_thinking_default = {
        role: spec.get("thinking_level")
        for role, spec in override_models.items()
        if role != "coordinator" and isinstance(spec, dict)
    }

    normalized_lookup: dict[str, Any] = {
        "llm_provider": cfg.get("llm_provider"),
        "llm_model": cfg.get("llm_model"),
        "coordinator_thinking": (
            cfg["coordinator_thinking"]
            if "coordinator_thinking" in cfg
            else coordinator_spec.get("thinking_level")
        ),
        "subagent_thinking": (
            cfg["subagent_thinking"]
            if "subagent_thinking" in cfg
            else subagent_thinking_default
        ),
        "embedding_mode": cfg.get("embedding_mode") or cfg.get("embedding_provider"),
        "embedding_model": cfg.get("embedding_model"),
        "embedding_dimension": cfg.get("embedding_dimension"),
        "document_converter_id": cfg.get("document_converter_id"),
        "docling_mode": cfg.get("docling_mode"),
        "datalab_merge_cross_page": cfg.get("datalab_merge_cross_page"),
        "citations_enabled": cfg.get("citations_enabled"),
        "references_section_enabled": cfg.get("references_section_enabled"),
        "auto_render_mermaid_diagrams": cfg.get("auto_render_mermaid_diagrams"),
        "enable_mcp_server": cfg.get("enable_mcp_server"),
        "citation_verification_enabled": cfg.get("citation_verification_enabled"),
        "citation_verification_provider": cfg.get("citation_verification_provider"),
        "citation_verification_model": cfg.get("citation_verification_model"),
        "citation_verification_threshold": cfg.get("citation_verification_threshold"),
        "adaptive_compaction_enabled": (
            cfg["adaptive_compaction_enabled"]
            if "adaptive_compaction_enabled" in cfg
            else cfg.get("enable_adaptive_compaction")
        ),
        "session_retention_days": cfg.get("session_retention_days"),
        "session_max_count": (
            cfg["session_max_count"]
            if "session_max_count" in cfg
            else cfg.get("max_saved_sessions")
        ),
    }

    result: dict[str, Any] = {}
    for key in sorted(ALLOWED_CONFIG_KEYS):
        val = normalized_lookup.get(key)
        if isinstance(val, str):
            val = sanitize_text(val, zotero_data_dir=resolved_dir)
        elif isinstance(val, dict):
            val = {
                str(k): sanitize_text(v, zotero_data_dir=resolved_dir)
                if isinstance(v, str)
                else v
                for k, v in val.items()
            }
        result[key] = val

    raw_base_url = cfg.get("llm_base_url")
    if isinstance(raw_base_url, str) and raw_base_url.strip():
        result["llm_base_url"] = sanitize_url(raw_base_url, zotero_data_dir=resolved_dir)

    return result

extract_shallow_turn_context 

extract_shallow_turn_context(session, *, invocation_id=None, error_message=None, include_prompts=False, include_item_titles=False, fallback_active_items=None, zotero_data_dir=None)

Extract shallow turn metadata and tool trace without chunk payloads or thoughts.

Source code in zocrates/utils/diagnostics.py 
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
def extract_shallow_turn_context(
    session: Any | None,
    *,
    invocation_id: str | None = None,
    error_message: str | None = None,
    include_prompts: bool = False,
    include_item_titles: bool = False,
    fallback_active_items: list[Any] | None = None,
    zotero_data_dir: Path | str | None = None,
) -> dict[str, Any]:
    """Extract shallow turn metadata and tool trace without chunk payloads or thoughts."""
    session_id = str(getattr(session, "id", "") or "") if session is not None else ""
    resolved_inv_id = sanitize_text(invocation_id, zotero_data_dir=zotero_data_dir) if invocation_id else None
    clean_error = (
        sanitize_text(error_message, zotero_data_dir=zotero_data_dir)
        if error_message
        else None
    )

    user_evt = None
    asst_evts: list[Any] = []

    if session is not None:
        events = list(getattr(session, "events", []) or [])
        has_rewinds = any(
            isinstance(
                getattr(getattr(e, "actions", None), "rewind_before_invocation_id", None),
                str,
            )
            for e in events
        )
        if has_rewinds:
            with contextlib.suppress(Exception):
                events = _apply_rewinds(events)
        events = [e for e in events if not _is_compaction_event(e)]

        turns: list[tuple[Any | None, list[Any]]] = []
        cur_user = None
        cur_asst: list[Any] = []
        for evt in events:
            author = getattr(evt, "author", "assistant")
            if author == "user":
                u_txt, _, _, u_frs = extract_event_parts(evt)
                if not u_txt.strip() and u_frs:
                    if cur_user is not None:
                        cur_asst.append(evt)
                    continue
                if cur_user is not None:
                    turns.append((cur_user, cur_asst))
                cur_user = evt
                cur_asst = []
            else:
                if cur_user is not None:
                    cur_asst.append(evt)
        if cur_user is not None:
            turns.append((cur_user, cur_asst))

        if turns:
            matched_turn = None
            if invocation_id:
                for u_e, a_es in reversed(turns):
                    if u_e is not None and getattr(u_e, "invocation_id", None) == invocation_id:
                        matched_turn = (u_e, a_es)
                        break
                    if any(getattr(ae, "invocation_id", None) == invocation_id for ae in a_es):
                        matched_turn = (u_e, a_es)
                        break
            if matched_turn is None and not invocation_id:
                matched_turn = turns[-1]
            if matched_turn is not None:
                user_evt, asst_evts = matched_turn
                if not resolved_inv_id:
                    cand_inv = getattr(user_evt, "invocation_id", None) or (
                        getattr(asst_evts[-1], "invocation_id", None) if asst_evts else None
                    )
                    if cand_inv:
                        resolved_inv_id = sanitize_text(str(cand_inv), zotero_data_dir=zotero_data_dir)

    # Context chips from user event state_delta or frontend fallback.
    raw_chips = None
    if user_evt is not None:
        actions = getattr(user_evt, "actions", None)
        state_delta = getattr(actions, "state_delta", {}) if actions else {}
        if isinstance(state_delta, dict):
            raw_chips = state_delta.get("active_context_items")
    if not raw_chips and fallback_active_items:
        raw_chips = fallback_active_items

    chips = _format_context_chips(
        raw_chips,
        include_item_titles=include_item_titles,
        zotero_data_dir=zotero_data_dir,
    )

    # Shallow tool trace: tool names, sorted argument keys ONLY, status, and sanitized error.
    # Exclude isolation_scope retrieval events, thoughts, and raw tool return payloads.
    trace_entries: list[dict[str, Any]] = []
    call_index_by_id: dict[str, int] = {}

    for evt in asst_evts:
        if getattr(evt, "isolation_scope", None) is not None:
            continue
        author_label = format_author_name(getattr(evt, "author", None))
        _, _thoughts, calls, responses = extract_event_parts(evt)

        for call in calls:
            tool_name = str(call.get("name") or "")
            if not tool_name or tool_name in ("finish_task", "adk_request_input"):
                continue
            raw_args = call.get("args")
            arg_keys = (
                sorted(str(k) for k in raw_args.keys())
                if isinstance(raw_args, dict)
                else []
            )
            entry: dict[str, Any] = {
                "tool": sanitize_text(tool_name, zotero_data_dir=zotero_data_dir),
                "author": author_label,
                "arg_keys": arg_keys,
                "status": "dispatched",
                "error": None,
            }
            call_id = call.get("id")
            if isinstance(call_id, str) and call_id:
                call_index_by_id[call_id] = len(trace_entries)
            trace_entries.append(entry)

        for resp in responses:
            tool_name = str(resp.get("name") or "")
            if not tool_name or tool_name in ("finish_task", "adk_request_input"):
                continue
            payload = resp.get("response")
            status = "ok"
            err_detail: str | None = None
            if isinstance(payload, dict):
                raw_err = payload.get("error")
                raw_status = str(payload.get("status") or "").lower()
                if raw_err:
                    status = "error"
                    err_detail = sanitize_text(
                        str(raw_err)[:300], zotero_data_dir=zotero_data_dir
                    )
                elif raw_status in ("error", "failed"):
                    status = "error"
                    msg_val = payload.get("message") or payload.get("detail") or ""
                    if msg_val:
                        err_detail = sanitize_text(
                            str(msg_val)[:300], zotero_data_dir=zotero_data_dir
                        )

            call_id = resp.get("id")
            if isinstance(call_id, str) and call_id in call_index_by_id:
                target = trace_entries[call_index_by_id[call_id]]
                target["status"] = status
                if err_detail:
                    target["error"] = err_detail
            else:
                trace_entries.append(
                    {
                        "tool": sanitize_text(tool_name, zotero_data_dir=zotero_data_dir),
                        "author": author_label,
                        "arg_keys": [],
                        "status": status,
                        "error": err_detail,
                    }
                )

    # Prompt & assistant text strictly gated by include_prompts.
    if include_prompts:
        user_prompt_text = ""
        if user_evt is not None:
            raw_u_txt, _, _, _ = extract_event_parts(user_evt)
            user_prompt_text = sanitize_text(raw_u_txt, zotero_data_dir=zotero_data_dir)

        asst_response_text = ""
        for evt in reversed(asst_evts):
            if getattr(evt, "isolation_scope", None) is not None:
                continue
            raw_a_txt, _, _, _ = extract_event_parts(evt)
            if raw_a_txt.strip():
                asst_response_text = sanitize_text(raw_a_txt, zotero_data_dir=zotero_data_dir)
                break
        if not asst_response_text and clean_error:
            asst_response_text = clean_error

        prompt_section: dict[str, Any] = {
            "included": True,
            "user_prompt": user_prompt_text or None,
            "assistant_response": asst_response_text or None,
        }
    else:
        prompt_section = {
            "included": False,
            "placeholder": OMITTED_PROMPTS_TEXT,
        }

    return {
        "session_id": sanitize_text(session_id, zotero_data_dir=zotero_data_dir) or None,
        "invocation_id": resolved_inv_id,
        "error_message": clean_error,
        "context_chips": chips,
        "prompts": prompt_section,
        "execution_trace": trace_entries,
    }

format_author_name 

format_author_name(raw_author)

Format internal agent names into customer-facing names without subagent references.

Source code in zocrates/utils/session_formatting.py 
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
def format_author_name(raw_author: str | None) -> str:
    """Format internal agent names into customer-facing names without subagent references."""
    if not raw_author or raw_author.lower() in ("user", "human"):
        return "User"
    s = raw_author.strip()
    if s.startswith("Zocrates/"):
        s = s[9:].strip()

    s = re.sub(r"sub[-_]?agent", "agent", s, flags=re.IGNORECASE)
    normalized = re.sub(r"[-_]", " ", s).lower().strip()
    if normalized in (
        "library synthesizer agent",
        "library synthesizer",
        "synthesizer agent",
        "synthesizer",
        "synthesis agent",
    ):
        return "Library Synthesis"
    if normalized in (
        "library search agent",
        "library search",
        "search agent",
        "search",
        "search subagent",
    ):
        return "Library Search"
    if normalized in ("online search agent", "online search", "discovery search agent"):
        return "Online Discovery"
    if normalized in ("online synthesizer agent", "online synthesizer", "discovery synthesizer agent"):
        return "Discovery Synthesis"
    if normalized in (
        "coordinator agent", "coordinator", "dialogue agent", "dialogue", "dialogue coordinator",
        "zocrates mode router", "zocrates mode router node", "mode router", "mode router node", "router",
    ):
        return "Zocrates"

    words = [w.capitalize() for w in re.split(r"[-_]+", s) if w]
    res = " ".join(words)
    res = re.sub(r"\bSubagent\b", "Agent", res, flags=re.IGNORECASE)
    return res or "Zocrates"

format_markdown_report 

format_markdown_report(data)

Format diagnostic snapshot into GitHub Markdown strictly capped at <= 50 KB.

Source code in zocrates/utils/diagnostics.py 
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
def format_markdown_report(data: dict[str, Any]) -> str:
    """Format diagnostic snapshot into GitHub Markdown strictly capped at <= 50 KB."""
    logs_info = data.get("logs") or {}
    orig_log_lines = list(logs_info.get("lines") or [])
    turn = data.get("turn_context") or {}
    orig_trace = list(turn.get("execution_trace") or [])

    rendered = _render_markdown_from_parts(
        data,
        log_lines=orig_log_lines,
        trace_entries=orig_trace,
    )
    if len(rendered.encode("utf-8")) <= MAX_REPORT_BYTES:
        return rendered

    # Step 1: Progressively trim older log lines first (preserving most recent tail).
    cur_logs = orig_log_lines
    logs_truncated = False
    if cur_logs:
        logs_truncated = True
        low, high = 0, len(cur_logs)
        best_count = 0
        while low <= high:
            mid = (low + high) // 2
            tail_slice = cur_logs[-mid:] if mid > 0 else []
            candidate = _render_markdown_from_parts(
                data,
                log_lines=tail_slice,
                trace_entries=orig_trace,
                logs_truncated=True,
            )
            if len(candidate.encode("utf-8")) <= MAX_REPORT_BYTES:
                best_count = mid
                low = mid + 1
            else:
                high = mid - 1
        cur_logs = cur_logs[-best_count:] if best_count > 0 else []
        rendered = _render_markdown_from_parts(
            data,
            log_lines=cur_logs,
            trace_entries=orig_trace,
            logs_truncated=True,
        )
        if len(rendered.encode("utf-8")) <= MAX_REPORT_BYTES:
            return rendered

    # Step 2: Trim execution trace entries if still exceeding 50 KB.
    cur_trace = orig_trace
    trace_truncated = False
    if cur_trace:
        trace_truncated = True
        for keep in (25, 10, 5, 0):
            if keep >= len(cur_trace):
                continue
            cur_trace = cur_trace[-keep:] if keep > 0 else []
            rendered = _render_markdown_from_parts(
                data,
                log_lines=cur_logs,
                trace_entries=cur_trace,
                logs_truncated=logs_truncated,
                trace_truncated=True,
            )
            if len(rendered.encode("utf-8")) <= MAX_REPORT_BYTES:
                return rendered

    # Step 3: Clamp prompt/response/error text lengths.
    for cap in (8000, 3000, 1000, 300):
        rendered = _render_markdown_from_parts(
            data,
            log_lines=cur_logs,
            trace_entries=cur_trace,
            max_prompt_chars=cap,
            logs_truncated=logs_truncated,
            trace_truncated=trace_truncated,
        )
        if len(rendered.encode("utf-8")) <= MAX_REPORT_BYTES:
            return rendered

    # Step 4: Unconditional UTF-8 byte ceiling guard.
    suffix = f"\n{TRUNCATION_MARKER}\n"
    suffix_bytes = suffix.encode("utf-8")
    encoded = rendered.encode("utf-8")
    prefix = encoded[: MAX_REPORT_BYTES - len(suffix_bytes)].decode(
        "utf-8", errors="ignore"
    )
    return prefix + suffix

get_auth_token 

get_auth_token(zotero_data_dir=None)

Resolve or generate the local backend authentication token.

Resolution order
  1. ZOCRATES_AUTH_TOKEN environment variable.
  2. Sidecar file <zocrates-data>/backend.auth.
  3. Cached in-memory token, or newly generated 256-bit hex token (secrets.token_hex(32)) persisted to backend.auth with mode 0o600.
Source code in zocrates/config.py 
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
def get_auth_token(zotero_data_dir: Path | str | None = None) -> str:
    """Resolve or generate the local backend authentication token.

    Resolution order:
      1. ``ZOCRATES_AUTH_TOKEN`` environment variable.
      2. Sidecar file ``<zocrates-data>/backend.auth``.
      3. Cached in-memory token, or newly generated 256-bit hex token
         (``secrets.token_hex(32)``) persisted to ``backend.auth`` with mode ``0o600``.
    """
    global _CACHED_AUTH_TOKEN

    env_token = os.getenv("ZOCRATES_AUTH_TOKEN", "").strip()
    if env_token:
        if zotero_data_dir is not None:
            auth_file = Path(zotero_data_dir) / ZOCRATES_DATA_SUBDIR / "backend.auth"
            if not auth_file.exists():
                _write_auth_sidecar(auth_file, env_token)
        return env_token

    if zotero_data_dir is not None:
        resolved_dir = Path(zotero_data_dir)
        auth_file = resolved_dir / ZOCRATES_DATA_SUBDIR / "backend.auth"
        try:
            if auth_file.is_file():
                file_token = auth_file.read_text(encoding="utf-8").strip()
                if file_token:
                    _CACHED_AUTH_TOKEN = file_token
                    return file_token
        except OSError:
            pass
        token = _CACHED_AUTH_TOKEN or secrets.token_hex(32)
        _write_auth_sidecar(auth_file, token)
        _CACHED_AUTH_TOKEN = token
        return token

    if _CACHED_AUTH_TOKEN:
        return _CACHED_AUTH_TOKEN

    resolved_dir = _resolve_explicit_zotero_data_dir(None)
    if resolved_dir is not None:
        auth_file = resolved_dir / ZOCRATES_DATA_SUBDIR / "backend.auth"
        try:
            if auth_file.is_file():
                file_token = auth_file.read_text(encoding="utf-8").strip()
                if file_token:
                    _CACHED_AUTH_TOKEN = file_token
                    return file_token
        except OSError:
            pass
        token = secrets.token_hex(32)
        _write_auth_sidecar(auth_file, token)
        _CACHED_AUTH_TOKEN = token
        return token

    _CACHED_AUTH_TOKEN = secrets.token_hex(32)
    return _CACHED_AUTH_TOKEN

get_datalab_api_key 

get_datalab_api_key(zotero_data_dir=None)

Return the Datalab API key from environment variables.

Zotero Preferences / Login Manager is the authoritative persistent vault for the Datalab API key. The backend stores the key in-memory (synced from Zotero via environment variable DATALAB_API_KEY on startup or POST /config at runtime).

Parameters:

Name Type Description Default
zotero_data_dir Path | None

Optional path to the Zotero data directory.

 None

Returns:

Type Description
str | None

The API key string or None if not configured.
Source code in zocrates/config.py 
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
def get_datalab_api_key(zotero_data_dir: Path | None = None) -> str | None:
    """Return the Datalab API key from environment variables.

    Zotero Preferences / Login Manager is the authoritative persistent vault for
    the Datalab API key. The backend stores the key in-memory (synced from Zotero
    via environment variable DATALAB_API_KEY on startup or POST /config at runtime).

    Args:
        zotero_data_dir: Optional path to the Zotero data directory.

    Returns:
        The API key string or None if not configured.
    """
    return get_secure_api_key(
        "datalab_api_key",
        env_vars=["DATALAB_API_KEY", "CHANDRA_API_KEY"],
    )

get_docling_hardware_assessment 

get_docling_hardware_assessment()

Assess system hardware capacity and recommend optimal local Docling configuration.

Returns:

Type Description
dict[str, Any]

Dict with keys recommended_mode ("fast" | "full"), is_gpu_accelerated (bool),
dict[str, Any]

vram_gb (float | None), and reason (str).
Source code in zocrates/rag/hardware.py 
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
def get_docling_hardware_assessment() -> dict[str, Any]:
    """Assess system hardware capacity and recommend optimal local Docling configuration.

    Returns:
        Dict with keys ``recommended_mode`` ("fast" | "full"), ``is_gpu_accelerated`` (bool),
        ``vram_gb`` (float | None), and ``reason`` (str).
    """
    info = get_system_hardware_info()
    high_perf = is_high_performance_gpu()

    if high_perf:
        vram_str = f"{info.vram_gb:.1f} GB VRAM" if info.vram_gb else f"{info.ram_gb:.1f} GB Unified Memory"
        return {
            "recommended_mode": "full",
            "is_gpu_accelerated": True,
            "vram_gb": info.vram_gb,
            "reason": f"High-performance accelerator detected ({info.gpu_name}, {vram_str}). Running in Full Mode (GPU layout, tables, and formula LaTeX VLM enabled).",
        }

    if info.gpu_vendor == "nvidia":
        if info.vram_gb is not None:
            reason = (
                f"GPU VRAM ({info.vram_gb:.1f} GB) is below the 6.0 GB threshold for heavy formula VLM. "
                f"Running in Fast Mode (layout & tables on CPU). For full LaTeX math OCR on this system, use Datalab Chandra (Cloud)."
            )
        else:
            reason = (
                f"Detected {info.gpu_name} with undetermined VRAM. "
                f"Running in Fast Mode (layout & tables on CPU). For full LaTeX math OCR on this system, use Datalab Chandra (Cloud)."
            )
        return {
            "recommended_mode": "fast",
            "is_gpu_accelerated": False,
            "vram_gb": info.vram_gb,
            "reason": reason,
        }

    if info.gpu_vendor == "apple":
        return {
            "recommended_mode": "fast",
            "is_gpu_accelerated": False,
            "vram_gb": None,
            "reason": f"Unified Memory ({info.ram_gb:.1f} GB) is below the 16.0 GB threshold. Running in Fast Mode (layout & tables on CPU).",
        }

    if info.gpu_vendor in ("amd", "intel"):
        vram_clause = f" ({info.vram_gb:.1f} GB VRAM)" if info.vram_gb is not None else ""
        return {
            "recommended_mode": "fast",
            "is_gpu_accelerated": False,
            "vram_gb": info.vram_gb,
            "reason": f"Detected {info.gpu_name}{vram_clause}. Running in Fast Mode (layout & tables on CPU). For full LaTeX math OCR on this system, use Datalab Chandra (Cloud).",
        }

    return {
        "recommended_mode": "fast",
        "is_gpu_accelerated": False,
        "vram_gb": None,
        "reason": f"Running on {info.gpu_name} (CPU). Running in Fast Mode (layout & tables on CPU). For full LaTeX math OCR on this system, use Datalab Chandra (Cloud).",
    }

get_openalex_api_key 

get_openalex_api_key(zotero_data_dir=None)

Retrieve the configured OpenAlex API key from the environment.

Source code in zocrates/config.py 
1308
1309
1310
1311
1312
1313
def get_openalex_api_key(zotero_data_dir: Path | None = None) -> str | None:
    """Retrieve the configured OpenAlex API key from the environment."""
    return get_secure_api_key(
        "openalex_api_key",
        env_vars=["OPENALEX_API_KEY"],
    )

get_sanitized_log_tail 

get_sanitized_log_tail(zotero_data_dir, max_lines=DEFAULT_LOG_TAIL_LINES)

Read and sanitize the tail of zocrates.log with zocrates.log.1 fallback.

Bounds line count between 50 and 500, sanitizes each line before length capping so partial secret prefixes cannot leak, and caps each line to 500 characters. Fails open (returns []) if the log directory or files do not exist.

Source code in zocrates/utils/diagnostics.py 
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
def get_sanitized_log_tail(
    zotero_data_dir: Path | str | None,
    max_lines: int = DEFAULT_LOG_TAIL_LINES,
) -> list[str]:
    """Read and sanitize the tail of ``zocrates.log`` with ``zocrates.log.1`` fallback.

    Bounds line count between 50 and 500, sanitizes each line before length capping
    so partial secret prefixes cannot leak, and caps each line to 500 characters.
    Fails open (returns ``[]``) if the log directory or files do not exist.
    """
    bounded_lines = max(MIN_LOG_TAIL_LINES, min(MAX_LOG_TAIL_LINES, int(max_lines)))
    resolved_dir = _resolve_zotero_dir_readonly(zotero_data_dir)
    if resolved_dir is None:
        return []

    log_dir = resolved_dir / ZOCRATES_DATA_SUBDIR / "logs"
    primary_log = log_dir / "zocrates.log"
    rotated_log = log_dir / "zocrates.log.1"

    raw_lines = _read_tail_lines_binary(primary_log, max_lines=bounded_lines)
    if len(raw_lines) < bounded_lines:
        needed = bounded_lines - len(raw_lines)
        prev_lines = _read_tail_lines_binary(rotated_log, max_lines=needed)
        if prev_lines:
            raw_lines = (prev_lines + raw_lines)[-bounded_lines:]

    sanitized_lines: list[str] = []
    exact_secrets = _collect_exact_secrets(resolved_dir)
    path_variants = _collect_path_variants(resolved_dir)
    for raw_line in raw_lines:
        # Bound pathological multi-megabyte lines before regex pass, then sanitize before 500-char cap.
        pre_bounded = raw_line[:4096]
        clean_line = sanitize_text(
            pre_bounded,
            zotero_data_dir=resolved_dir,
            exact_secrets=exact_secrets,
            path_variants=path_variants,
        )
        if len(clean_line) > MAX_LOG_LINE_CHARS:
            clean_line = clean_line[: MAX_LOG_LINE_CHARS - 3] + "..."
        sanitized_lines.append(clean_line)

    return sanitized_lines

get_secure_api_key 

get_secure_api_key(key_name='openrouter_api_key', env_vars=('OPENROUTER_API_KEY', 'OPEN_ROUTER_API_KEY'))

Load a named API key from in-memory session environment variables.

Zotero Preferences / Login Manager serves as the secure persistent vault. Backend keys are in-memory session variables (os.environ) synced from Zotero.

Parameters:

Name Type Description Default
key_name str

Entry identifier (e.g., 'openrouter_api_key', 'openai_api_key').

 'openrouter_api_key'
env_vars list[str] | tuple[str, ...] | None

Environment variable names to check.

 ('OPENROUTER_API_KEY', 'OPEN_ROUTER_API_KEY')

Returns:

Type Description
str | None

The API key string or None if not found.
Source code in zocrates/config.py 
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
def get_secure_api_key(
    key_name: str = "openrouter_api_key",
    env_vars: list[str] | tuple[str, ...] | None = ("OPENROUTER_API_KEY", "OPEN_ROUTER_API_KEY"),
) -> str | None:
    """Load a named API key from in-memory session environment variables.

    Zotero Preferences / Login Manager serves as the secure persistent vault.
    Backend keys are in-memory session variables (os.environ) synced from Zotero.

    Args:
        key_name: Entry identifier (e.g., 'openrouter_api_key', 'openai_api_key').
        env_vars: Environment variable names to check.

    Returns:
        The API key string or None if not found.
    """
    if env_vars:
        for var in env_vars:
            val = os.getenv(var)
            if val:
                return val
    return None

get_system_hardware_info 

get_system_hardware_info(*, force_refresh=False)

Probe system memory and GPU accelerator information.

Caches result in memory to prevent repeated subprocess calls during server runtime.

Checks, in priority order: Apple Silicon MPS, NVIDIA via nvidia-smi or PyTorch / procfs / sysfs fallback, AMD via rocm-smi or sysfs fallback, Intel via xpu-smi or sysfs fallback, then falls back to CPU-only.

Returns:

Type Description
SystemHardwareInfo

A SystemHardwareInfo dataclass populated with RAM, GPU vendor,
SystemHardwareInfo

GPU name, VRAM, and compute capability (None for non-NVIDIA).
Source code in zocrates/rag/hardware.py 
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
def get_system_hardware_info(*, force_refresh: bool = False) -> SystemHardwareInfo:
    """Probe system memory and GPU accelerator information.

    Caches result in memory to prevent repeated subprocess calls during server runtime.

    Checks, in priority order: Apple Silicon MPS, NVIDIA via ``nvidia-smi`` or
    PyTorch / procfs / sysfs fallback, AMD via ``rocm-smi`` or sysfs fallback,
    Intel via ``xpu-smi`` or sysfs fallback, then falls back to CPU-only.

    Returns:
        A ``SystemHardwareInfo`` dataclass populated with RAM, GPU vendor,
        GPU name, VRAM, and compute capability (``None`` for non-NVIDIA).
    """
    global _CACHED_HW_INFO
    if _CACHED_HW_INFO is not None and not force_refresh:
        return _CACHED_HW_INFO

    ram_gb = get_system_ram_gb()

    # 1. Apple Silicon (macOS MPS)
    if sys.platform == "darwin" and platform.machine() in ("arm64", "aarch64"):
        _CACHED_HW_INFO = SystemHardwareInfo(
            ram_gb=ram_gb,
            gpu_vendor="apple",
            gpu_name="Apple Silicon (MPS)",
            vram_gb=ram_gb,  # Unified Memory Architecture
            compute_capability=None,
        )
        return _CACHED_HW_INFO

    # 2. NVIDIA Detection
    # 2A: Try nvidia-smi CLI
    try:
        output = subprocess.check_output(
            [
                "nvidia-smi",
                "--query-gpu=compute_cap,name,memory.total",
                "--format=csv,noheader,nounits",
            ],
            text=True,
            stderr=subprocess.DEVNULL,
        ).strip()
        if output:
            line = output.splitlines()[0]
            parts = [p.strip() for p in line.split(",")]
            cc = float(parts[0])
            gpu_name = parts[1] if len(parts) > 1 else "NVIDIA GPU"
            vram_mb = float(parts[2]) if len(parts) > 2 else 0.0
            _CACHED_HW_INFO = SystemHardwareInfo(
                ram_gb=ram_gb,
                gpu_vendor="nvidia",
                gpu_name=gpu_name,
                vram_gb=vram_mb / 1024.0 if vram_mb else None,
                compute_capability=cc,
            )
            return _CACHED_HW_INFO
    except Exception:
        pass

    # 2B: Fallback via PyTorch / Procfs / Sysfs
    torch_info = _detect_nvidia_torch()
    if torch_info is not None:
        gpu_name, vram_gb, cc = torch_info
        _CACHED_HW_INFO = SystemHardwareInfo(
            ram_gb=ram_gb,
            gpu_vendor="nvidia",
            gpu_name=gpu_name,
            vram_gb=vram_gb,
            compute_capability=cc,
        )
        return _CACHED_HW_INFO

    procfs_model = _detect_nvidia_procfs()
    pci_found = _detect_nvidia_pci()
    if procfs_model or pci_found:
        _CACHED_HW_INFO = SystemHardwareInfo(
            ram_gb=ram_gb,
            gpu_vendor="nvidia",
            gpu_name=procfs_model or "NVIDIA GPU",
            vram_gb=None,
            compute_capability=None,
        )
        return _CACHED_HW_INFO

    # 3. AMD Detection
    # 3A: Try rocm-smi CLI
    try:
        output = subprocess.check_output(
            ["rocm-smi", "--showid"],
            text=True,
            stderr=subprocess.DEVNULL,
        ).strip()
        if output:
            _CACHED_HW_INFO = SystemHardwareInfo(
                ram_gb=ram_gb,
                gpu_vendor="amd",
                gpu_name="AMD Radeon / Instinct (ROCm)",
                vram_gb=None,
                compute_capability=None,
            )
            return _CACHED_HW_INFO
    except Exception:
        pass

    # 3B: Fallback via Linux Sysfs
    amd_sysfs = _detect_amd_sysfs()
    if amd_sysfs is not None:
        amd_name, amd_vram = amd_sysfs
        _CACHED_HW_INFO = SystemHardwareInfo(
            ram_gb=ram_gb,
            gpu_vendor="amd",
            gpu_name=amd_name,
            vram_gb=amd_vram,
            compute_capability=None,
        )
        return _CACHED_HW_INFO

    # 4. Intel Detection
    # 4A: Try xpu-smi CLI
    try:
        output = subprocess.check_output(
            ["xpu-smi", "health"],
            text=True,
            stderr=subprocess.DEVNULL,
        ).strip()
        if output:
            _CACHED_HW_INFO = SystemHardwareInfo(
                ram_gb=ram_gb,
                gpu_vendor="intel",
                gpu_name="Intel Arc / Data Center GPU (XPU)",
                vram_gb=None,
                compute_capability=None,
            )
            return _CACHED_HW_INFO
    except Exception:
        pass

    # 4B: Sysfs fallback for Intel discrete / Arc / iGPU
    intel_sysfs = _detect_intel_sysfs()
    if intel_sysfs is not None:
        intel_name, intel_vram = intel_sysfs
        _CACHED_HW_INFO = SystemHardwareInfo(
            ram_gb=ram_gb,
            gpu_vendor="intel",
            gpu_name=intel_name,
            vram_gb=intel_vram,
            compute_capability=None,
        )
        return _CACHED_HW_INFO

    # 5. CPU Fallback
    _CACHED_HW_INFO = SystemHardwareInfo(
        ram_gb=ram_gb,
        gpu_vendor="cpu",
        gpu_name=platform.processor() or "CPU",
        vram_gb=None,
        compute_capability=None,
    )
    return _CACHED_HW_INFO

resolve_zotero_data_dir 

resolve_zotero_data_dir(zotero_data_dir=None)

Resolve the active Zotero data directory with auto-discovery fallback.

Parameters:

Name Type Description Default
zotero_data_dir Path | str | None

Explicit data directory path or None.

 None

Returns:

Type Description
Path | None

Resolved Path to the Zotero data directory, or None if not located.
Source code in zocrates/config.py 
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
def resolve_zotero_data_dir(zotero_data_dir: Path | str | None = None) -> Path | None:
    """Resolve the active Zotero data directory with auto-discovery fallback.

    Args:
        zotero_data_dir: Explicit data directory path or None.

    Returns:
        Resolved Path to the Zotero data directory, or None if not located.
    """
    if zotero_data_dir is not None:
        return Path(zotero_data_dir)
    try:
        from zocrates.rag.service_manager import get_syncer
        syncer = get_syncer()
        if syncer and getattr(syncer, "_zotero_data_dir", None):
            return Path(syncer._zotero_data_dir)
    except Exception:
        pass
    if ZOTERO_DATA_DIR:
        return Path(ZOTERO_DATA_DIR)
    if env_dir := os.getenv("ZOTERO_DATA_DIR"):
        return Path(env_dir)
    # Auto-discover active Zotero profile in user home directory (fallback for offline scripts)
    candidates = (
        list(Path.home().glob("Zotero*/zocrates-data"))
        + list(Path.home().glob("Zotero*/*/zocrates-data"))
        + list(Path.home().glob(".var/app/org.zotero.Zotero/data/zotero/zocrates-data"))
        + list(Path.home().glob(".zotero/zotero/*/zocrates-data"))
    )
    if candidates:
        selected = candidates[0].parent
        if len(candidates) > 1:
            logger.warning(
                "Multiple Zotero profiles found with zocrates-data (%s). "
                "Defaulting to %s. Set ZOTERO_DATA_DIR to select a specific profile.",
                [str(c.parent) for c in candidates],
                selected,
            )
        else:
            logger.debug("Auto-discovered Zotero data directory: %s", selected)
        return selected
    return None

sanitize_text 

sanitize_text(text, zotero_data_dir=None, *, exact_secrets=None, path_variants=None)

Scrub secrets, tokens, emails, and multi-OS filesystem paths from text.

Applies two independent layers
  1. Exact-value replacement for live auth tokens, API keys, zotero_data_dir, and Path.home().
  2. Pre-compiled regex scrubbing for OpenRouter/LLM keys, Bearer/Auth headers, query tokens, URL credentials, emails, and Linux/macOS/Windows/UNC paths.
Source code in zocrates/utils/diagnostics.py 
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
def sanitize_text(
    text: str | None,
    zotero_data_dir: Path | str | None = None,
    *,
    exact_secrets: list[tuple[str, str]] | None = None,
    path_variants: list[tuple[str, str]] | None = None,
) -> str:
    """Scrub secrets, tokens, emails, and multi-OS filesystem paths from ``text``.

    Applies two independent layers:
      1. Exact-value replacement for live auth tokens, API keys, ``zotero_data_dir``,
         and ``Path.home()``.
      2. Pre-compiled regex scrubbing for OpenRouter/LLM keys, Bearer/Auth headers,
         query tokens, URL credentials, emails, and Linux/macOS/Windows/UNC paths.
    """
    if text is None:
        return ""
    out = str(text)
    if not out:
        return ""

    resolved_dir = _resolve_zotero_dir_readonly(zotero_data_dir)

    # Layer 1A: Exact secret replacement.
    secrets_list = exact_secrets if exact_secrets is not None else _collect_exact_secrets(resolved_dir)
    for secret, placeholder in secrets_list:
        if secret in out:
            out = out.replace(secret, placeholder)
        url_encoded = urllib.parse.quote(secret, safe="")
        if url_encoded != secret and len(url_encoded) >= 6 and url_encoded in out:
            out = out.replace(url_encoded, placeholder)

    # Layer 1B: Exact filesystem root replacement (<ZOTERO_DATA_DIR> before <HOME>).
    variants_list = path_variants if path_variants is not None else _collect_path_variants(resolved_dir)
    for variant, placeholder in variants_list:
        if variant in out:
            out = out.replace(variant, placeholder)

    # Layer 2: Pattern-based scrubbing.
    out = _OPENROUTER_KEY_RE.sub("[REDACTED_OPENROUTER_KEY]", out)
    out = _GENERIC_SK_KEY_RE.sub("[REDACTED_API_KEY]", out)
    out = _BEARER_TOKEN_RE.sub(r"\1[REDACTED_TOKEN]", out)
    out = _ZOCRATES_AUTH_HEADER_RE.sub(r"\1[REDACTED_AUTH_TOKEN]", out)
    out = _X_API_KEY_HEADER_RE.sub(r"\1[REDACTED_API_KEY]", out)
    out = _QUERY_TOKEN_RE.sub(r"\1[REDACTED_TOKEN]", out)
    out = _URL_USERINFO_RE.sub(r"\1[REDACTED_CREDENTIALS]@", out)
    out = _MAILTO_RE.sub("mailto:[REDACTED_EMAIL]", out)
    out = _EMAIL_RE.sub("[REDACTED_EMAIL]", out)

    # Multi-OS user paths & Windows UNC paths.
    out = _LINUX_VAR_HOME_RE.sub("<HOME>", out)
    out = _LINUX_HOME_RE.sub("<HOME>", out)
    out = _MACOS_USERS_RE.sub("<HOME>", out)
    out = _WIN_ESCAPED_USERS_RE.sub("<HOME>", out)
    out = _WIN_BACKSLASH_USERS_RE.sub("<HOME>", out)
    out = _WIN_SLASH_USERS_RE.sub("<HOME>", out)
    out = _WIN_ESCAPED_UNC_RE.sub("<UNC_PATH>", out)
    out = _WIN_UNC_RE.sub("<UNC_PATH>", out)

    return out

sanitize_url 

sanitize_url(url, zotero_data_dir=None)

Strip userinfo credentials, query parameters, and fragments from a URL.

Source code in zocrates/utils/diagnostics.py 
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
def sanitize_url(url: str | None, zotero_data_dir: Path | str | None = None) -> str:
    """Strip userinfo credentials, query parameters, and fragments from a URL."""
    if not url or not str(url).strip():
        return ""
    raw = str(url).strip()
    try:
        parsed = urllib.parse.urlsplit(raw)
        if not parsed.scheme and not parsed.netloc:
            return sanitize_text(raw, zotero_data_dir=zotero_data_dir)
        host = parsed.hostname or ""
        if ":" in host and not host.startswith("["):
            host = f"[{host}]"
        try:
            port = parsed.port
        except ValueError:
            port = None
        clean_netloc = f"{host}:{port}" if port is not None else host
        cleaned = urllib.parse.urlunsplit((parsed.scheme, clean_netloc, parsed.path, "", ""))
        return sanitize_text(cleaned, zotero_data_dir=zotero_data_dir)
    except Exception:
        return sanitize_text(raw, zotero_data_dir=zotero_data_dir)