context-brain

Showing 25 of 32 from the live probe.

• Query Memory

  [stable] Search Context Brain semantic memory and return a grounded answer with citations. Use for direct factual recall about known projects, past decisions, test results, architecture, or known bugs. If confidence is low, fall back to search_raw_chunks for graph-aware retrieval.

• Search Raw Chunks

  [stable] Retrieve raw knowledge chunks with semantic distance scores and graph/hub-aware retrieval signals. Use when reasoning, comparing, synthesizing across multiple sources, or when query_memory returns low confidence. Returns chunk_text plus optional graph_match, hub_match, knowledge_path, retrieval_reason, ranking_reason, and score_components for each result. Treat hub signals as relevance/navigation signals — factual claims must come from chunk_text. Set debug=True to

• Health Check

  [stable] Check whether Context Brain backend is online and responsive.

• List Hubs

  [stable] List all knowledge hubs (user-defined or AI-suggested navigation entry points). Hubs are NOT domains — they are entry points that group related content for retrieval.

• Get Hub

  [stable] Get details for a specific knowledge hub including aliases, related terms, and linked content.

• Create Hub

  [stable] Create a new knowledge hub (navigation entry point). Hubs group related content under a queryable navigation node — they are NOT domains. Args: title: Hub display name (required). hub_type: Hub classification — must pass this exact parameter name. One of: 'project', 'system', 'concept_cluster', 'topic'. Default: 'topic'. description: Optional longer description. aliases: Optional list of alternative names for this hub. relate

• Update Hub

  [stable] Update an existing knowledge hub. Only provided fields are changed. status: 'active' or 'archived'.

• Link Content To Hub

  [stable] Link a stored content item (by content_id) to a knowledge hub. Use to attach previously saved memory to a hub for hub-aware retrieval boosting.

• Save And Link To Hub

  [stable] Save content to Context Brain and immediately link it to a hub in one call. Replaces the 3-step (save → search → link) dance needed when using save_memory + link_content_to_hub separately. Returns content_id on success. Args: content: The text to save (200+ words recommended for reliable retrieval). save_purpose: Why this is being saved — intent or context (required). hub_id: Target hub UUID to link the saved content to (required). content_u

• List Hubs Json

  [experimental] List all knowledge hubs as structured JSON with count. Returns {"hubs": [...], "count": N} where each hub has id, title, type, status, aliases, related_terms. Use this instead of list_hubs when you need machine-readable data (e.g. for graph rendering, dedup checks, edge creation).

• Get Hub Json

  [experimental] Get a knowledge hub as structured JSON with linked content and stats. Returns {"hub": {...}, "linked_content": [...], "stats": {linked_content_count, edge_count}}. Use when you need hub detail for edge creation, graph inspection, or content audit.

• Get Graph Snapshot

  [experimental] Get full graph snapshot: all hub nodes + curated edges. Returns {"nodes": [...], "edges": [...], "stats": {...}, "schema_version": "v1"}. Nodes are active hubs. Edges are persisted curated relationships (manual, approved, rejected). Inferred edges are computed in the Live Artifact — not returned here. Use to bootstrap the knowledge map, audit coverage, or plan curation work.

• List Graph Snapshot

  [experimental] Alias for get_graph_snapshot — spec-canonical name. Get full graph snapshot: all hub nodes + curated edges. Returns {"nodes": [...], "edges": [...], "stats": {...}, "schema_version": "v1"}.

• List Hub Edges

  [experimental] List persisted hub-to-hub edges (curated relationships). Optionally filter to edges involving a specific hub_id. By default excludes rejected and archived edges. Returns {"edges": [...], "count": N}. Edge fields: id, source_hub_id, target_hub_id, type, status, origin, confidence, reason, note, edge_key, created_by, created_at, updated_at.

• Create Hub Edge

  [experimental] Create a manual hub-to-hub edge in the knowledge graph. edge_type must be one of: parent, related, duplicate, overlaps, supports, part_of, contradicts. Symmetric types (related, duplicate, overlaps, contradicts) are deduplicated regardless of direction. Returns {"edge": {...}, "created": true} or error if active edge already exists. Use to persist a relationship between two hubs that you have identified as meaningful.

• Update Hub Edge

  [experimental] Update an existing hub edge. Only provided fields are changed. status: manual | approved | rejected | needs_review | archived edge_type: parent | related | duplicate | overlaps | supports | part_of | contradicts Returns {"edge": {...}, "updated": true}.

• Archive Hub Edge

  [experimental] Archive a hub edge. Archived edges are excluded from the active graph. Archiving frees the edge_key so a new edge of the same type can be created. Returns {"archived": true, "edge": {...}}.

• Approve Inferred Edge

  [experimental] Approve an inferred hub-to-hub edge and persist it to the backend. Use when the Live Artifact or your own analysis identifies a valid inferred edge that should be preserved across sessions. Idempotent — safe to call multiple times. Returns {"edge": {...}, "approved": true}.

• Reject Inferred Edge

  [experimental] Reject an inferred hub-to-hub edge and persist the rejection to the backend. Use when the Live Artifact suggests an edge that is incorrect or misleading. Persisting the rejection prevents the same inferred edge from being re-suggested. Idempotent — safe to call multiple times. Returns {"edge": {...}, "rejected": true}.

• Search Graph Preview

  [stable] Lightweight metadata-first search — no full chunk text returned. Use this BEFORE load_graph_node_full to navigate the knowledge graph cheaply. Each result shows: content_id, node_type, quick_summary, hub_match, score. load_full_content_recommended=true means the node has high relevance but no quick_summary, so loading the full content is needed to judge relevance. Workflow: search_graph_preview → decide which nodes to load → load_graph_node_full. Args:

• Load Graph Node Full

  [stable] Load full content text plus all Phase 4/5 metadata for a specific node. Use after search_graph_preview to retrieve the complete document for nodes you decided are relevant. Returns full_text, node_type, quick_summary, domain, word_count, and created_at in one call. Args: content_id: UUID of the content item (e.g. from search_graph_preview results).

• Get Content By Id

  [stable] Retrieve full content text for a known content_id. Use this when search_raw_chunks returns a truncated chunk_text and you need the complete original document. Returns the full stored text plus basic metadata (domain, word_count, created_at). Args: content_id: UUID of the content item (e.g. 'fe90eee3-7a38-42cb-8031-bd276eba280b')

• Create Decision Memory

  [experimental] Save a decision with its reasoning, context, and supporting references. Use after important architectural, product, or strategic decisions. Stores not just *what* was decided but *why* — enabling future agents to retrieve the full rationale and avoid repeating past deliberations. If this decision supersedes an older one, pass its UUID in supersedes_decision_id — the old decision is automatically marked superseded and linked back. Args: title: Sho

• Explain Decision

  [experimental] Retrieve full decision record including reasoning, context, and why it matters. Use to recall the rationale behind a past decision before making a related choice, or to surface the decision for review/update. Args: decision_id: UUID of the decision to retrieve.

• List Decisions

  [experimental] List decisions, optionally filtered by status or hub. Args: status: Filter by status — active, superseded, reversed, or draft. hub_id: Filter to decisions linked to this hub UUID. limit: Max results to return (default 20).