Workhorse-tier batch proof checking. Verify logical proofs or claims in batch.

Deep-tier change brief. Generate an evidence-first brief about code changes.

Deep-tier change pack. Run a deterministic pipeline to analyze code changes and write a durable artifact.

Last-resort workhorse-tier chat. Free-form conversation with the model.

Instant-tier classifier. Assign a label from a closed set to a text input. Returns the label, confidence, and reasoning.

Instant-tier code citation. Cite specific code locations for claims.

Instant-tier code map. Generate a structural map of a codebase.

Deep-tier code review. Perform a comprehensive code review.

Artifact-tier corpus amendment. Manually add, remove, or modify chunks in a corpus.

Artifact-tier corpus amendment history. View the history of amendments to a corpus.

Deep-tier corpus-grounded answering. Answer a question using a corpus as the knowledge base.

Artifact-tier corpus health check. Verify the integrity and health of an indexed corpus.

Embed-tier corpus indexing. Build or update a named corpus on disk. Idempotent: unchanged files are reused from the existing corpus (by sha256); changed files are re-embedded; removed files are dropped from the corpus.

Artifact-tier corpus listing. List all indexed corpora and their metadata.

Embed-tier corpus refresh. Re-embed an existing corpus to update embeddings.

Embed-tier corpus reranking. Rerank search results using a reranking model.

Embed-tier corpus search. Search an indexed corpus for relevant chunks.

Artifact-tier diagnostic tool. Check server health, Ollama connectivity, model availability, and configuration.

Workhorse-tier draft generator. Generate prose, code, or structured content from a prompt.

Embed-tier embedding. Convert text to a vector embedding using a local embedding model.

Embed-tier semantic search. Search a corpus of embeddings for semantically similar text.

Instant-tier structured extraction. Extract entities, relationships, or fields from text into a schema.

Workhorse-tier hypothesis drilling. Test hypotheses against code and logs.

Deep-tier incident brief. Generate an evidence-first incident brief from logs and source files.

Deep-tier incident pack. Run a deterministic pipeline to investigate an incident and write a durable artifact.

Artifact-tier log statistics. Compute statistics over the NDJSON event log.

Artifact-tier log tail. Retrieve the last N entries from the NDJSON event log.

Workhorse-tier multi-file refactor proposal. Propose refactoring changes across multiple files.

Workhorse-tier refactor planning. Plan a refactoring strategy.

Deep-tier repository brief. Generate an evidence-first brief about a repository.

Deep-tier repository pack. Run a deterministic pipeline to analyze a repository and write a durable artifact.

FLAGSHIP. Answer a question grounded in specific files. Takes FILE PATHS (not raw text) — reads and chunks locally, returns a digest with validated citations. Use this to understand a repo/doc without burning Claude context on the full content. Citations outside source_paths are stripped server-side; cited line_ranges that point past EOF have the range dropped (path is kept) with a warning. Returns honest grounding signals: `weak: true` when an answer has zero validated citations (likely ungrounded); `abstain` when the question is outside the corpus.

Deep-tier deep summarizer. Summarize text or files with depth and nuance prioritized.

Workhorse-tier fast summarizer. Summarize text or files with speed prioritized over depth.

Instant-tier log triage. Parse structured or unstructured logs, extract error patterns, and return a summary with severity levels.

Workhorse-tier claim verification. Verify claims against source material.