Semantic search over the knowledge base.

Searches markdown documents with YAML frontmatter. Understands Russian and English. Use expand_graph=true to include knowledge graph neighbors (structurally related docs).

Args: query: Search query (e.g. "privacy architecture", "API design patterns") n_results: Number of results (default 5) doc_type: Filter by type (depends on your KB schema) expand_graph: Expand results with knowledge graph neighbors (1-hop)

Semantic search over Claude Code chat session history.

Finds past sessions by what was discussed or worked on. Useful for "how did I solve X?" or "when did I work on Y?" questions.

Args: query: Search query (e.g. "knowledge graph implementation", "OCR receipt scanning") n_results: Number of results (default 5) project: Filter by project name (e.g. "my-app", "backend")

Execute a raw Cypher query against the code intelligence graph.

The graph contains: Project, File, Symbol, Package, Session nodes. Edges: HAS_FILE, DEFINES, DEPENDS_ON, MODIFIED, IN_PROJECT, TOUCHED, EDITED, CREATED, IMPORTS (File->File or File->Package), CALLS (File->Symbol), INHERITS (Symbol->Symbol).

Example queries:

• MATCH (p:Project) RETURN p.name, p.path LIMIT 10

• MATCH (f:File {project: 'my-app'}) RETURN f.path, f.lang LIMIT 20

• MATCH (p:Project)-[:DEPENDS_ON]->(pkg:Package) WHERE pkg.name = 'react' RETURN p.name

• MATCH (s:Session)-[:EDITED]->(f:File) RETURN f.path, COUNT(s) ORDER BY COUNT(s) DESC LIMIT 10

Args: query: Cypher query string

Get code intelligence graph statistics.

Returns counts of projects, files, symbols, packages, sessions, and edge type breakdown.

Architecture overview of a project from the code graph.

Returns structured data: stack, languages, directory layers, key patterns (mixins, base classes, CRUD schemas), top dependencies, and hub files.

Args: project: Project name (e.g. "my-app", "backend-api")

Generate a YAML-formatted repository map showing the most important files and their symbols.

Helps understand the global structure of a project without reading all files.

Args: project: Project name (e.g. "my-app") limit: Maximum number of top files to include (default 20)

Packages shared across multiple projects in the code graph.

Returns list of packages with the projects that depend on them, sorted by number of projects (most shared first). Useful for finding common dependencies and potential shared infrastructure.

Get project information from the registry.

Without name: returns list of all active projects with stacks. With name: returns detailed info for one project.

Args: name: Project name (e.g. "my-app"). Omit for full list.

Fetch a URL with browser-like headers to bypass basic bot protection.

Returns page content as plain text (HTML tags stripped) or raw HTML. Handles redirects, cookies, and common anti-bot headers automatically.

Args: url: URL to fetch max_length: Max content length in chars (default 20000) extract_text: Strip HTML tags and return plain text (default true) timeout: Request timeout in seconds (default 30)

Semantic search over project source code and documentation.

Searches indexed project codebases (per-project FalkorDB vector DBs). Auto-indexes the project on first search if no index exists. Useful for finding code patterns, functions, classes, and docs across projects.

Args: query: Search query (e.g. "authentication middleware", "API route handler") n_results: Number of results (default 5) project: Project name or path (e.g. "my-app", "~/projects/my-app"). Omit to search all. chunk_type: Filter by "code" or "doc"

Reindex a project's source code and docs into FalkorDB vectors.

Call this after significant code changes to update the search index. Uses sentence-transformers backend (safe for memory).

Args: project: Project name or path (e.g. "my-app", "~/projects/my-app")

Search indexed external sources (Telegram, YouTube, etc.).

Each source is stored in its own FalkorDB graph. YouTube videos are chunked by chapters — results include chapter name and timecode. Without source filter, searches all sources and merges by relevance.

Args: query: Search query (e.g. "startup idea", "revenue growth") source: Filter by source name (e.g. "telegram", "youtube"). Omit to search all. n_results: Number of results (default 5)

List all auto-detected topics with video counts.

Tags are assigned automatically via zero-shot embedding similarity during video indexing. Shared across videos — enables topic clustering.

Args: source: Source name (default: "youtube")

Find related videos by shared tags.

Given a video URL, finds other videos that share auto-detected topics. Returns videos sorted by number of shared tags (most related first).

Args: video_url: YouTube video URL or video ID source: Source name (default: "youtube")

List indexed external sources with document counts.

Shows all source graphs under ~/.solo/sources/ with their sizes. YouTube sources include video/chunk breakdown (videos, video_chunks fields).