chezmoi_lib.sh | Shared shell library for merge/apply hooks (source selection, atomic writes, checksums, LiteLLM base) |
verify_templates.py | Render every chezmoi *.tmpl via execute-template to catch errors early |
verify_mermaids.py | Check .mermaids/ file-census counts against the effective git file set (part of make check) |
verify_bin_surface.py | Check comma-command completions, docs tokens, catalog tokens, and library ownership |
verify_docs_navigation.py | Check docs/reference/ links and implementation coverage catalog rows |
verify_agent_file_sizes.py | Check agent skill reference markdown stays under the 20 KB harness view limit (SKILL.md exempt) |
yaml_parser.py | Minimal dependency-free YAML parser for project data files |
jsonc_dump.py | JSONC serializer matching OpenCode's trailing-comma config style |
mcp_registry.py | Read/normalize the canonical MCP registry |
generate_mcp_configs.py | Generate tool-specific MCP configs from mcp_servers.yaml |
inject_mcp_into_codex_toml.py | Inject MCP servers into Codex TOML at a marker line |
inject_mcp_into_opencode_jsonc.py | Inject MCP servers into an OpenCode JSONC placeholder |
merge_claude_mcp.py | Surgically update only mcpServers in ~/.claude.json |
merge_opencode_models.py | Merge LiteLLM/Azure models into OpenCode JSONC |
merge_copilot_settings.py | Typed declared-over-live merge for Copilot settings.json (exact subagents.agents) |
managed_config_manifest.py | Declared-key manifests shared by the config merge hooks |
generated_artifact_ledger.py | Record/evaluate generated artifacts for semantic drift (backs ,doctor ai) |
ai_models.py | Parse the litellm_models / azure_models sections of ai_models.yaml |
model_mirrors.py | Generate/verify the committed model-mirror JSON + Go artifacts from the registry sources |
model_mirror_consumer.py | Stdlib fail-closed consumer views of the model mirror for deployed commands such as ,ai |
model_capabilities.v1.json | Hand-verified harness capability snapshot consumed by model-mirror generation |
generate_pi_models.py | Build Pi models.json from the shared base plus LiteLLM/Azure providers |
model_display.py | Shared display-name formatting for LiteLLM model entries |
probe_litellm_prompt_cache.py | Probe repeated-prompt and tool-schema cache signals across LiteLLM models |
analyze_pi_session.py | Report privacy-safe token/cache/compaction/re-read metrics from one Pi v3 session |
ai_kb.py | Local markdown + SQLite FTS5/vector knowledge base for agent runs |
embed.py | Stdlib embedding-service abstraction: one-shot embed_runner.py runs plus connect-only resident dispatch via embed_client.py |
embed_runner.py | Isolated PEP 723 fastembed runner (BAAI/bge-small-en-v1.5, 384-d) |
embed_client.py | Deadline-bounded unix-socket client for the resident embed worker (ensure/ping/embed) |
embed_worker.py | PEP 723 resident fastembed worker serving embeddings over a private unix socket |
vec_runner.py | Isolated PEP 723 sqlite-vec KNN/pairs runner for the KB |
agent_memory.py | Inspect/wipe hook memory under /tmp/specs for the current workspace |
worklog_queue.py | Crash-safe bounded per-session worklog event queue flushed into /tmp/specs topic worklogs |
sync_llama_cpp_models.py | Download missing GGUF files declared in the llama.cpp manifest |
reconcile_golang_pkgs.py | Reconcile Go binaries against the manifest (hook 05-update-golang-pkgs) |
reconcile_custom_packages.py | Reconcile source-installed custom package artifacts |
yt_search.py | YouTube search backend (powers ,youtube-search) |
tests/ | Python test suite (test_scripts.py, agent-hook tests) |