Mechanical catalog
The Living Index
Every room, exhibit and source span in one place. Cross references come from PHAROS's recorded graph, or are written comparisons and say so.
Pass 3 of 3 · Deep dive of the hallPass 1 · OverviewPass 2 · How it works
Text catalog
Every entry, as plain links
- The Hall The central rooms and the ways out of them.
- PHAROS The Infinite Catalog ROOM BUILT
- The book: contents
- Pass 1 · Overview
- Pass 2 · How it works
- What’s the idea, in one picture?
- Exhibit 1: Which repositories has PHAROS really indexed? RECORDED
- Exhibit 2: What will PHAROS refuse to index? RUNS HERE
- Exhibit 3: What does it read from a repository? RECORDED
- Exhibit 4: How does it cut code into pieces? RECORDED
- Exhibit 5: What does it turn into vectors, and what does indexing cost? RUNS HERE
- Exhibit 6: How does it know who calls whom? RECORDED
- Exhibit 7: Which retriever answered, and why that one? RUNS HERE
- Exhibit 8: What if the code changed after it was indexed? RUNS HERE
- Exhibit 9: What does an assistant get back? RECORDED
- How do the parts connect?
- Pass 3 · Deep dive
- The system
- Ingestion jobs SOURCE AUDITED
- Claim a durable job PINNED SOURCE
- Indexer SOURCE AUDITED
- Index one commit PINNED SOURCE
- Diff by blob SHA PINNED SOURCE
- Parsers SOURCE AUDITED
- Pick a parser PINNED SOURCE
- Embeddings SOURCE AUDITED
- Pick the embedder PINNED SOURCE
- Embed the question PINNED SOURCE
- What gets embedded PINNED SOURCE
- Call graph HAS A RECORDED RUN
- Who calls this, and what does it call? RECORDED
- Resolve calls conservatively PINNED SOURCE
- Match a name PINNED SOURCE
- Read the neighbours PINNED SOURCE
- PostgreSQL store SOURCE AUDITED
- The two search indexes PINNED SOURCE
- Retriever HAS A RECORDED RUN
- Hints turn auto into hybrid RECORDED
- A missing embedder: degrade with a warning, or refuse RECORDED
- Choose a mode, then run it PINNED SOURCE
- Is dense possible here? PINNED SOURCE
- Dense search with pgvector PINNED SOURCE
- Reciprocal rank fusion PINNED SOURCE
- Exact symbol boost PINNED SOURCE
- Down-rank tests and docs PINNED SOURCE
- Match a repository reference PINNED SOURCE
- Source verification HAS A RECORDED RUN
- A question reaches verified source RECORDED
- What the blob check does with a mismatch RECORDED
- The Git blob hash PINNED SOURCE
- Compare before returning PINNED SOURCE
- Load at the pinned commit PINNED SOURCE
- Fetch from an allowlisted host PINNED SOURCE
- Resolve pointers concurrently PINNED SOURCE
- Pointer to request PINNED SOURCE
- Context assembly HAS A RECORDED RUN
- A question reaches verified source RECORDED
- Assemble the context PINNED SOURCE
- Render with citations PINNED SOURCE
- The character budget PINNED SOURCE
- Resolve the repository PINNED SOURCE
- One hop of the graph PINNED SOURCE
- Repository memory FAILED ITS GATE
- Memory enters the context PINNED SOURCE
- MCP server HAS A RECORDED RUN
- A question reaches verified source RECORDED
- Who calls this, and what does it call? RECORDED
- The MCP tool PINNED SOURCE
- The find_symbol tool PINNED SOURCE
- Constant-time token check PINNED SOURCE
- What the read token may do PINNED SOURCE
- REST API HAS A RECORDED RUN
- Hints turn auto into hybrid RECORDED
- The REST search endpoint PINNED SOURCE
- What the read token may do PINNED SOURCE
- Attach pointers to hits PINNED SOURCE
- Ingestion jobs SOURCE AUDITED
- Symbol shelves
- codeintel/__init__.py
- codeintel SOURCE-DERIVED
- codeintel/context.py
- context SOURCE-DERIVED
- RepositoryNotFound SOURCE-DERIVED
- ContextRequest SOURCE-DERIVED
- _Budget SOURCE-DERIVED
- take SOURCE-DERIVED
- resolve_repositories SOURCE-DERIVED
- hit_to_dict SOURCE-DERIVED
- source_requests SOURCE-DERIVED
- assemble_context SOURCE-DERIVED
- select_rules SOURCE-DERIVED
- _render_memory SOURCE-DERIVED
- _cite SOURCE-DERIVED
- _render SOURCE-DERIVED
- codeintel/describe.py
- describe SOURCE-DERIVED
- DescribeStats SOURCE-DERIVED
- grounded SOURCE-DERIVED
- _candidates SOURCE-DERIVED
- describe_repository SOURCE-DERIVED
- codeintel/edge_server.py
- edge_server SOURCE-DERIVED
- _Bucket SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- take SOURCE-DERIVED
- make_handler SOURCE-DERIVED
- main SOURCE-DERIVED
- codeintel/embeddings.py
- embeddings SOURCE-DERIVED
- EmbeddingUnavailable SOURCE-DERIVED
- _normalize SOURCE-DERIVED
- _pad SOURCE-DERIVED
- Embedder SOURCE-DERIVED
- embed_documents SOURCE-DERIVED
- embed_query SOURCE-DERIVED
- encode_raw SOURCE-DERIVED
- describe SOURCE-DERIVED
- HashEmbedder SOURCE-DERIVED
- _features SOURCE-DERIVED
- embed_documents SOURCE-DERIVED
- _use_prefixes SOURCE-DERIVED
- _PrefixMixin SOURCE-DERIVED
- _doc SOURCE-DERIVED
- _query SOURCE-DERIVED
- SentenceTransformerEmbedder SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- _load SOURCE-DERIVED
- _encode SOURCE-DERIVED
- embed_documents SOURCE-DERIVED
- embed_query SOURCE-DERIVED
- encode_raw SOURCE-DERIVED
- HttpEmbedder SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- _post SOURCE-DERIVED
- _embed SOURCE-DERIVED
- embed_documents SOURCE-DERIVED
- embed_query SOURCE-DERIVED
- OnnxEmbedder SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- _load SOURCE-DERIVED
- embed_documents SOURCE-DERIVED
- embed_query SOURCE-DERIVED
- encode_raw SOURCE-DERIVED
- EdgeFirstEmbedder SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- _http SOURCE-DERIVED
- _reset_http SOURCE-DERIVED
- _edge SOURCE-DERIVED
- encode_raw SOURCE-DERIVED
- embed_documents SOURCE-DERIVED
- embed_query SOURCE-DERIVED
- describe SOURCE-DERIVED
- _read_profile SOURCE-DERIVED
- resolve_backend SOURCE-DERIVED
- get_embedder SOURCE-DERIVED
- reset_embedder_cache SOURCE-DERIVED
- codeintel/git_source.py
- git_source SOURCE-DERIVED
- RepoSpecError SOURCE-DERIVED
- GitError SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- RepoRef SOURCE-DERIVED
- is_github SOURCE-DERIVED
- raw_url SOURCE-DERIVED
- web_url SOURCE-DERIVED
- raw_file_url SOURCE-DERIVED
- web_file_url SOURCE-DERIVED
- parse_repo_spec SOURCE-DERIVED
- _is_ip_literal SOURCE-DERIVED
- _local_ref SOURCE-DERIVED
- validate_ref SOURCE-DERIVED
- _git_env SOURCE-DERIVED
- _sanitize SOURCE-DERIVED
- _classify SOURCE-DERIVED
- _run SOURCE-DERIVED
- _auth_config SOURCE-DERIVED
- TreeEntry SOURCE-DERIVED
- Checkout SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- __enter__ SOURCE-DERIVED
- __exit__ SOURCE-DERIVED
- close SOURCE-DERIVED
- _git SOURCE-DERIVED
- list_files SOURCE-DERIVED
- read_blob SOURCE-DERIVED
- open_checkout SOURCE-DERIVED
- read_file_at_commit SOURCE-DERIVED
- local_git_dir SOURCE-DERIVED
- codeintel/graph.py
- graph SOURCE-DERIVED
- ChunkNode SOURCE-DERIVED
- Edge SOURCE-DERIVED
- GraphStats SOURCE-DERIVED
- as_dict SOURCE-DERIVED
- _python_module_index SOURCE-DERIVED
- _resolve_python_import SOURCE-DERIVED
- _resolve_path_import SOURCE-DERIVED
- build_edges SOURCE-DERIVED
- codeintel/indexer.py
- indexer SOURCE-DERIVED
- LanguageStats SOURCE-DERIVED
- IndexStats SOURCE-DERIVED
- lang SOURCE-DERIVED
- as_dict SOURCE-DERIVED
- _utcnow SOURCE-DERIVED
- _decode SOURCE-DERIVED
- _looks_minified SOURCE-DERIVED
- _mean_vector SOURCE-DERIVED
- _PreparedChunk SOURCE-DERIVED
- CodeIndexer SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- get_or_create_repository SOURCE-DERIVED
- index SOURCE-DERIVED
- _sync SOURCE-DERIVED
- _skip_reason SOURCE-DERIVED
- _delete_resources SOURCE-DERIVED
- _repoint_unchanged SOURCE-DERIVED
- _index_batch SOURCE-DERIVED
- _build_rows SOURCE-DERIVED
- _embed SOURCE-DERIVED
- _rebuild_graph SOURCE-DERIVED
- _count_chunks SOURCE-DERIVED
- _count_lines SOURCE-DERIVED
- codeintel/jobs.py
- jobs SOURCE-DERIVED
- utcnow SOURCE-DERIVED
- default_worker_id SOURCE-DERIVED
- Submission SOURCE-DERIVED
- QueueFull SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- submit SOURCE-DERIVED
- claim SOURCE-DERIVED
- next_queued_ids SOURCE-DERIVED
- retry_delay_seconds SOURCE-DERIVED
- heartbeat SOURCE-DERIVED
- cancel SOURCE-DERIVED
- queue_summary SOURCE-DERIVED
- recover_stale SOURCE-DERIVED
- classify_error SOURCE-DERIVED
- run SOURCE-DERIVED
- codeintel/languages.py
- languages SOURCE-DERIVED
- LanguageSpec SOURCE-DERIVED
- detect_language SOURCE-DERIVED
- indexable_extensions SOURCE-DERIVED
- codeintel/notify.py
- notify SOURCE-DERIVED
- Notifier SOURCE-DERIVED
- notify SOURCE-DERIVED
- wait SOURCE-DERIVED
- _parse SOURCE-DERIVED
- RedisNotifier SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- notify SOURCE-DERIVED
- wait SOURCE-DERIVED
- UpstashRestNotifier SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- _cmd SOURCE-DERIVED
- notify SOURCE-DERIVED
- wait SOURCE-DERIVED
- get_notifier SOURCE-DERIVED
- parsers/__init__.py
- parsers SOURCE-DERIVED
- _module_path SOURCE-DERIVED
- line_chunks SOURCE-DERIVED
- parse_file SOURCE-DERIVED
- parsers/base.py
- base SOURCE-DERIVED
- Symbol SOURCE-DERIVED
- add_call SOURCE-DERIVED
- add_identifier SOURCE-DERIVED
- ParseResult SOURCE-DERIVED
- clip SOURCE-DERIVED
- parsers/python_parser.py
- python_parser SOURCE-DERIVED
- module_path_for SOURCE-DERIVED
- _resolve_relative SOURCE-DERIVED
- _signature SOURCE-DERIVED
- _docstring SOURCE-DERIVED
- _start_line SOURCE-DERIVED
- _Extractor SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- current SOURCE-DERIVED
- _enter SOURCE-DERIVED
- visit_ClassDef SOURCE-DERIVED
- _visit_func SOURCE-DERIVED
- visit_Call SOURCE-DERIVED
- visit_Name SOURCE-DERIVED
- visit_Attribute SOURCE-DERIVED
- visit_Import SOURCE-DERIVED
- visit_ImportFrom SOURCE-DERIVED
- parse_python SOURCE-DERIVED
- parsers/treesitter_parser.py
- treesitter_parser SOURCE-DERIVED
- _Def SOURCE-DERIVED
- _Lang SOURCE-DERIVED
- _text SOURCE-DERIVED
- _field_text SOURCE-DERIVED
- _last_segment SOURCE-DERIVED
- _callee SOURCE-DERIVED
- _c_declarator_name SOURCE-DERIVED
- _named_with_body SOURCE-DERIVED
- _js_variable_function SOURCE-DERIVED
- _go_type_spec SOURCE-DERIVED
- _rust_impl_name SOURCE-DERIVED
- _rust_impl_type SOURCE-DERIVED
- _strip_quotes SOURCE-DERIVED
- _js_require SOURCE-DERIVED
- _string_field SOURCE-DERIVED
- _get_parser SOURCE-DERIVED
- grammar_status SOURCE-DERIVED
- _clean_comment SOURCE-DERIVED
- _leading_comment SOURCE-DERIVED
- _end_line SOURCE-DERIVED
- _signature SOURCE-DERIVED
- _go_receiver_type SOURCE-DERIVED
- parse_treesitter SOURCE-DERIVED
- _file_comment SOURCE-DERIVED
- codeintel/retrieval.py
- retrieval SOURCE-DERIVED
- ModeUnavailable SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- Filters SOURCE-DERIVED
- Evidence SOURCE-DERIVED
- as_dict SOURCE-DERIVED
- Hit SOURCE-DERIVED
- SearchOutcome SOURCE-DERIVED
- is_test_path SOURCE-DERIVED
- path_role SOURCE-DERIVED
- wanted_roles SOURCE-DERIVED
- symbol_like_terms SOURCE-DERIVED
- compound_name SOURCE-DERIVED
- Retriever SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- search SOURCE-DERIVED
- _dense_unavailable_reason SOURCE-DERIVED
- _filter_clauses SOURCE-DERIVED
- _keyword SOURCE-DERIVED
- _keyword_python SOURCE-DERIVED
- _dense SOURCE-DERIVED
- _fuse SOURCE-DERIVED
- _symbol_boost SOURCE-DERIVED
- _expand_graph SOURCE-DERIVED
- _finalize SOURCE-DERIVED
- _hydrate SOURCE-DERIVED
- neighbors SOURCE-DERIVED
- _supports_iterative_scan SOURCE-DERIVED
- _cosine SOURCE-DERIVED
- repositories_matching SOURCE-DERIVED
- codeintel/runtime.py
- runtime SOURCE-DERIVED
- ensure_database SOURCE-DERIVED
- session_scope SOURCE-DERIVED
- secret SOURCE-DERIVED
- job_ref SOURCE-DERIVED
- run_options SOURCE-DERIVED
- codeintel/source.py
- source SOURCE-DERIVED
- SourceRequest SOURCE-DERIVED
- SourceResult SOURCE-DERIVED
- _FileCache SOURCE-DERIVED
- __init__ SOURCE-DERIVED
- get SOURCE-DERIVED
- put SOURCE-DERIVED
- git_blob_sha SOURCE-DERIVED
- _slice SOURCE-DERIVED
- _fetch_https SOURCE-DERIVED
- _load SOURCE-DERIVED
- _resolve_one SOURCE-DERIVED
- resolve_sources SOURCE-DERIVED
- codeintel/symbols.py
- symbols SOURCE-DERIVED
- find_symbol SOURCE-DERIVED
- symbol_graph SOURCE-DERIVED
- chunk_ref SOURCE-DERIVED
- repository_summary SOURCE-DERIVED
- hover SOURCE-DERIVED
- codeintel/text.py
- text SOURCE-DERIVED
- split_identifier SOURCE-DERIVED
- tokenize SOURCE-DERIVED
- query_terms SOURCE-DERIVED
- body_excerpt SOURCE-DERIVED
- _after_signature SOURCE-DERIVED
- with_description SOURCE-DERIVED
- semantic_summary SOURCE-DERIVED
- search_text SOURCE-DERIVED
- tsquery_from_terms SOURCE-DERIVED
- code/router.py
- router SOURCE-DERIVED
- job_out SOURCE-DERIVED
- _mode_unavailable SOURCE-DERIVED
- submit_repository SOURCE-DERIVED
- list_repositories SOURCE-DERIVED
- _get_repo SOURCE-DERIVED
- get_repository SOURCE-DERIVED
- delete_repository SOURCE-DERIVED
- list_jobs SOURCE-DERIVED
- get_job SOURCE-DERIVED
- cancel_job SOURCE-DERIVED
- _filters SOURCE-DERIVED
- search_code SOURCE-DERIVED
- retrieve_context SOURCE-DERIVED
- lookup_symbol SOURCE-DERIVED
- system_status SOURCE-DERIVED
- _repo_ids SOURCE-DERIVED
- _learning_job_out SOURCE-DERIVED
- queue_learning SOURCE-DERIVED
- list_learning_jobs SOURCE-DERIVED
- memory_brief SOURCE-DERIVED
- memory_search SOURCE-DERIVED
- memory_items SOURCE-DERIVED
- review_memory_item SOURCE-DERIVED
- memory_skills SOURCE-DERIVED
- memory_skill SOURCE-DERIVED
- mcp/native_server.py
- native_server SOURCE-DERIVED
- _error SOURCE-DERIVED
- _transport_security SOURCE-DERIVED
- build_mcp_server SOURCE-DERIVED
- get_streamable_http_app SOURCE-DERIVED
- shared/auth.py
- auth SOURCE-DERIVED
- Principal SOURCE-DERIVED
- is_admin SOURCE-DERIVED
- user_uuid SOURCE-DERIVED
- as_state SOURCE-DERIVED
- _configured SOURCE-DERIVED
- principal_for_token SOURCE-DERIVED
- parse_bearer SOURCE-DERIVED
- is_public SOURCE-DERIVED
- allowed_for SOURCE-DERIVED
- require_admin SOURCE-DERIVED
- token_problems SOURCE-DERIVED
- codeintel/__init__.py
- Revision journals
- Similarity moves into pgvector SOURCE-LINKED
- The knowledge-management product is removed SOURCE-LINKED
- Module boundaries become a CI gate SOURCE-LINKED
- A real MCP server SOURCE-LINKED
- Today's core: chunk-level indexing and hybrid retrieval SOURCE-LINKED
- Migrations squashed, code index schema added SOURCE-LINKED
- Two tokens, and working MCP on the new core SOURCE-LINKED
- Retrieval gets a benchmark SOURCE-LINKED
- Free deployment with CPU embeddings SOURCE-LINKED
- Granite embeddings and dense by default SOURCE-LINKED
- Query embedding can run off the host SOURCE-LINKED
- The memory evaluation is locked first SOURCE-LINKED
- Memory fails its locked evaluation SOURCE-LINKED
- Memory claims must prove themselves SOURCE-LINKED
- The second evaluation stops with no verdict SOURCE-LINKED
- The book: contents
- CognitiveOS / Valerie The Keeper's Study ROOM BUILT
- The book: contents
- Pass 1 · Overview
- Pass 2 · How it works
- What’s the idea, in one picture?
- Station 1: What does a model actually hand over? RUNS HERE
- Station 2: What is she allowed to point at? RUNS HERE
- Station 3: Can I throw the switch myself? RUNS HERE
- Station 4: What does a refusal do? RUNS HERE
- Station 5: How do I know what really happened? RUNS HERE
- Station 6: What happens in one turn on the Mac? SIMULATED
- Station 7: Would the Mac app do this? SIMULATED
- Station 8: What happens to text from outside? SIMULATED
- Station 9: What can she do on a Mac? SIMULATED
- How do the parts connect?
- Pass 3 · Deep dive
- The book: contents
- R&Bot The Noir Recording Chamber CONCEPT WING
- BABEL The Theatre of Stories CONCEPT WING
- The Electric Gallery A concept gallery CONCEPT WING
- The Mechanical Foundry A concept workshop CONCEPT WING
Cross references
| From | Relation | To |
|---|---|---|
| Ingestion jobs | runs | Indexer |
| Indexer | parses with | Parsers |
| Indexer | embeds with | Embeddings |
| Indexer | rebuilds | Call graph |
| Parsers | chunks | PostgreSQL store |
| Embeddings | vectors | PostgreSQL store |
| Call graph | edges | PostgreSQL store |
| PostgreSQL store | searched by | Retriever |
| Retriever | ranked hits | Context assembly |
| Source verification | verified code | Context assembly |
| Repository memory | memory block | Context assembly |
| Context assembly | retrieve_context | MCP server |
| Context assembly | /context/retrieve | REST API |
| Retriever | search_code | MCP server |
| Retriever | /search/code | REST API |
| Projects | Comparison | Reason |
|---|---|---|
| PHAROS, CognitiveOS / Valerie | Two control boundaries | PHAROS compares source with the Git blob recorded at indexing. A mismatch withholds the code, but with no recorded blob the code is returned marked unverified. Valerie checks typed moves against issued state before native engines run them. Both checks are deterministic code, not a model’s judgement. |
| PHAROS, CognitiveOS / Valerie | Two kinds of memory | PHAROS keeps repository evidence and conventions. Valerie keeps sourced temporal facts and a ledger. |
| BABEL, CognitiveOS / Valerie | Typed output | BABEL validates chapters against a typed schema on its batch path, but its single-chapter path accepts looser blocks, so validation is inconsistent. Valerie validates typed moves against issued state before native engines run them. |
Symbol-level calls and imports are on each symbol page, with direction and resolution.
