Distillery

Apply a pre-computed classification to an existing entry. USE WHEN: you have determined an entry's type and confidence via LLM or heuristic analysis and want to persist the classification result.

Read or update a runtime configuration value. USE WHEN: reading current thresholds/settings, or adjusting them at runtime without editing the config file directly.

Store a correction that supersedes an existing entry. USE WHEN: an existing entry contains wrong information and you want to replace it with corrected content while preserving the audit trail.

Find stored entries similar to the given text (cosine similarity). USE WHEN: checking for duplicates or conflicts before storing, finding entries related to arbitrary text, or surfacing hidden connections to a known entry (entries that are similar but not yet linked via relations). Supports progressive disclosure modes.

Retrieve a single knowledge entry by its unique ID. USE WHEN: fetching the full content and metadata of a specific entry (e.g. after finding its ID via search or list).

Sync GitHub issues and PRs into the knowledge base using a batched pipeline. url: repository slug (owner/repo) or full GitHub URL. author: author field for created entries (default: gh-sync). project: optional project name to scope entries. background: when true, runs async and returns a job_id immediately.

Ingest an arbitrary document (ADR, spec, decision, customer feedback). USE WHEN: importing a standalone document — a markdown ADR/spec/RFC, a design decision, or a customer-feedback transcript/doc — so it becomes queryable knowledge with provenance. Distinct from distillery_store (single entry, semantic dedup) and from PreCompact transcripts: this chunks large text into multiple linked entries and deduplicates idempotently by content hash, so re-ingesting identical content adds no second entry.

List knowledge entries with optional filters and pagination (newest first). USE WHEN: browsing or filtering entries without a semantic query. Use distillery_search instead when you have a natural-language question. By default, only entries with status in (active, pending_review) are returned — archived entries are hidden. Pass ``status="archived"`` to list only archived entries, ``status="any"`` to include every status, or ``include_archived=true`` to add archived entries to the default view.

Manage typed relations between knowledge entries. USE WHEN: linking entries together (e.g. marking one as blocking another, citing a reference, or flagging duplicates), walking the relation graph from a seed entry to surface multi-hop neighbours, or computing graph metrics (bridges, communities) on the relations subgraph.

Resolve a pending-review entry by approving, reclassifying, or archiving it. USE WHEN: acting on entries in the review queue (entries with status=pending_review from low-confidence classifications).

Search knowledge entries using semantic similarity (cosine distance, ranked descending). USE WHEN: finding entries that match a natural-language question or topic. Each result includes a similarity score (0-1, higher is more relevant). By default, only entries with status in (active, pending_review) are considered — archived entries are hidden. Pass ``status="archived"`` to search only archived entries, ``status="any"`` to include every status, or ``include_archived=true`` to add archived entries to the default candidate set. When ``expand_graph=true``, after the semantic search returns its seed result set, the tool BFS-expands 1 or 2 hops via ``entry_relations`` to surface structurally connected entries. Graph entries are scored at ``parent_score * 0.5 ** depth``, marked with ``provenance="graph"``, and merged into the result list (sorted by descending score, truncated to ``limit``). Seeds are tagged ``provenance="search"``. The envelope gains a ``graph_expansion`` summary. When ``expand_graph=false`` (default), the existing envelope is unchanged — strictly additive.

Return a lightweight in-protocol health/metadata probe. USE WHEN: verifying MCP connectivity (e.g. from the ``/setup`` wizard) without relying on the HTTP-only ``/health`` endpoint. Works uniformly on stdio and HTTP transports. PARAMS: (none) RETURNS (success): { status: "ok", version: str, # distillery package version build_sha: str, # git SHA (or "dev") transport: "stdio" | "http" | "unknown", tool_count: int, # number of registered MCP tools store: { entry_count: int | null, db_size_bytes: int | null }, embedding_provider: str, # model name or provider class name last_feed_poll: { source_count: int, last_poll_at: str | null }, uptime_seconds?: int # seconds since server startup } RELATED: distillery_list (for entry counts, filtering, and per-group aggregates), distillery_configure (to inspect/adjust runtime configuration)

Store a new knowledge entry and return its ID with dedup/conflict information. USE WHEN: capturing a new piece of knowledge (session notes, bookmarks, meeting minutes, ideas, etc.) into the Distillery store.

Batch-store multiple knowledge entries in one call (no dedup/conflict checks). USE WHEN: bulk-importing entries (e.g. GitHub history sync, migration, backfill) where per-entry dedup is unnecessary and throughput matters.

Check the status of background sync jobs. job_id: look up a specific job by ID. source_url: list jobs for a specific source URL. If neither is provided, lists all recent jobs.

Update one or more fields on an existing knowledge entry. USE WHEN: modifying an entry's content, type, tags, status, or other mutable fields. At least one updatable field must be provided.

Manage monitored feed sources for ambient intelligence. USE WHEN: listing, adding, or removing RSS/GitHub feed sources that Distillery polls for new content.