Skip to content

Latest commit

 

History

History
292 lines (210 loc) · 8.32 KB

File metadata and controls

292 lines (210 loc) · 8.32 KB

CLI Reference

Every KnowCode command, its flags, defaults, and examples. This page is the single source of truth for the command surface — other docs link here rather than restating flags or defaults.

KnowCode has 16 top-level commands: install, analyze, build, index, query, semantic-search, context, export, stats, telemetry, doctor, preflight, server, history, ask, and mcp-server.

Commands that require optional extras fail fast with an actionable hint (e.g., Install knowcode[server] to use 'knowcode server'.). The extras matrix and knowcode install are covered in Configuration.


Lifecycle commands

install

Install the dependencies for a full-featured KnowCode setup (equivalent to the search, llm, server, watch, mcp, and voyageai extras). Useful after a minimal pip install knowcode.

knowcode install [--upgrade] [--user] [--dry-run]
Flag Effect
--upgrade Upgrade packages while installing
--user Install into the Python user site-packages directory
--dry-run Print the installer command without running it

analyze

Scan and parse a directory to build the knowledge store (semantic graph of entities and relationships). Graph only — no semantic index.

knowcode analyze <directory> [--output <path>] [--ignore <pattern>] [--temporal] [--coverage <cobertura.xml>]
  • --temporal — also extract git authorship/change history (enables the history command and authored/changed_by relationships).
  • --coverage — merge runtime coverage signals from a Cobertura XML report.

Example:

knowcode analyze src/ --ignore "tests/*" --temporal

build

Build the knowledge base and the semantic index in one atomic step. Recommended first command for any new repository. Prints entity/relationship counts and the preflight report card; a failed build leaves the previous generation in place.

knowcode build [<directory>] [--ignore <pattern>] [--config <path>] [--temporal] [--incremental]

<directory> is optional and defaults to the current directory.

Example:

knowcode build . --ignore "tests/*"

index

Build (or update) the semantic search index — chunks, embeddings, BM25 tokens, vector store — published as part of a generation.

knowcode index <directory> [--output <path>] [--config <path>] [--incremental]

preflight

Grade the target codebase across 10 quality dimensions (documentation density, naming, parse success rate, …) and print a report card with recommendations. Cheap: it reuses the parsed graph, never re-reads sources. Also run automatically as part of build.

knowcode preflight [<directory>] [--ignore <pattern>] [--config <path>] [--json]

<directory> is optional and defaults to the current directory.

doctor

Check whether the local setup is ready for daily use: strict config loading, required API keys, knowledge store schema, semantic index schema and embedding dimensions, artifact disk footprint (default threshold 500 MB), unsupported-language warnings, freshness, and optionally a live MCP stdio handshake.

knowcode doctor [--store <path>] [--index <path>] [--config <path>] [--max-disk-mb <n>] [--mcp] [--json]

Example:

knowcode doctor --store . --mcp

Query commands

query

Lexical queries against the knowledge store.

knowcode query <type> <target> [--store <path>] [--json]

Query types:

Type Purpose
search <pattern> Search entities by name
callers <entity> Find what calls this entity
callees <entity> Find what this entity calls
deps <entity> Get all dependencies

Example:

knowcode query callers "GraphBuilder.build_from_directory"

semantic-search

Natural-language search against the semantic index (hybrid BM25 + vectors).

knowcode semantic-search <query> [--index <path>] [--store <path>] [--config <path>] [--limit <n>]

--limit defaults to 5.

Example:

knowcode semantic-search "Where is the graph built?"

context

Generate a token-budgeted context bundle for one entity, ready for AI consumption. Prints the bundle, its token count, and a truncation flag.

knowcode context <entity> [--store <path>] [--max-tokens <n>]

--max-tokens defaults to 2000.

Example:

knowcode context "GraphBuilder.build_from_directory" --max-tokens 4000

ask

Ask a natural-language question about the codebase. Classifies the query, retrieves the minimal sufficient context, and answers using the first available configured LLM (with failover). Requires knowcode[llm] and an API key; the store and index must already exist.

knowcode ask <question> [--store <path>] [--config <path>]

Example:

knowcode ask "How does the graph builder work?"

history

Show git history for the whole codebase or for one entity. Requires the store to have been built with --temporal.

knowcode history [target] [--store <path>] [--limit <n>]

--limit defaults to 10.

Example:

knowcode history "KnowledgeStore"

Output commands

export

Export the knowledge store as deterministic, no-LLM multi-level Markdown documentation: an index, an architecture overview, one page per module, and a documentation_manifest.json of entity content hashes for freshness checks.

knowcode export [--store <path>] [--output <dir>]

--output defaults to docs.

stats

Print entity and relationship counts by kind for the current store.

knowcode stats [--store <path>]

Serving commands

server

Start the local FastAPI intelligence server — the preferred surface for locally hosted AI agents. Binds 127.0.0.1:8000 by default (loopback only; there is no authentication). Requires knowcode[server]; --watch requires knowcode[watch].

knowcode server [--host <host>] [--port <port>] [--store <path>] [--watch]

Example:

knowcode server --port 8080 --watch

With --watch, modified/created files are queued for incremental re-indexing and deleted/renamed files have their chunks invalidated. The endpoint reference is in REST API.

mcp-server

Start the MCP (Model Context Protocol) server over stdio for IDE agent integration. Read-only and deterministic: tools never auto-run analysis.

knowcode mcp-server [--store <path>] [--config <path>]

Tools exposed: search_codebase, get_entity_context, trace_calls, retrieve_context_for_query, assess_codebase_quality. Client setup for Claude Desktop, VS Code, and Antigravity is covered in IDE & Agent Integration; the canonical retrieval policy is the MCP contract.

Telemetry

telemetry show / telemetry clear

Summarize or delete the local telemetry log. Telemetry never leaves your machine and never contains your questions or code — see Telemetry & Privacy.

knowcode telemetry show [--store <path>]
knowcode telemetry clear [--store <path>] [--yes]

Supported language matrix

Extension Language Parser mechanism Notes
.py Python Python AST Full semantic parsing (Python 3.10–3.12)
.js, .jsx JavaScript Tree-sitter Classes, functions, imports, calls, exports
.ts, .tsx TypeScript Tree-sitter Classes, functions, imports, calls, exports
.java Java Tree-sitter Classes, interfaces, methods, imports, inheritance
.rs Rust Tree-sitter Structs, enums, functions, impl blocks
.vue Vue Custom SFC scanner Vue Single-File Components (tree-sitter for embedded scripts)
.md Markdown Custom parser Heading hierarchy and document structure
.rst reStructuredText Custom parser Section structure via the shared RST heading scanner
.yaml, .yml YAML Custom parser Configuration keys with nested structure

Any extension not listed (e.g., .go, .cpp, .h, .swift, .rb, .php, .css, .html) is ignored during analyze/index/build. Per-parser construct coverage (what exactly each parser extracts, and its known limitations) is tracked in the parser construct matrix.