[Catalog](https://skills.polygres.com/index.md) / [polygres-sdk](https://skills.polygres.com/polygres-sdk/SKILL.md)

Published skills release: 0.9.0. [Compatibility and source](https://skills.polygres.com/manifest.json).

Scripts and templates mentioned here require the installed package; reading this page does not install them.

# MCP grounded Context answers

Read [`mcp-tool-contract.md`](https://skills.polygres.com/polygres-sdk/references/mcp-tool-contract.md) first. Use this playbook for an interactive grounded
answer when compatible MCP Context tools are available. Use the SDK references
for persistent Python application code.

1. Call `whoami` and resolve the project boundary.
2. Inspect `get_context_collection` and
   `get_context_collection_status`; use `verify_context_collection` when current
   serving evidence is required.
3. Read the collection's default vector, dimensions, metric, text inputs,
   result columns, filters, and source key contract.
4. Inspect `get_context_capabilities` for the chosen method and
   `query_embedding_generation` when using text. For a vector with a saved
   embedding configuration, pass `text` to the existing Context search tool.
   Polygres uses its configured model. For application-supplied vectors,
   construct a query vector with the original model and dimensions.
5. Choose exactly one `context_search`, `search_full_text`,
   `context_text_hybrid_search`, `context_graph_first_search`,
   `context_first_graph_search`, `context_rank_fusion_search`, or
   `context_joint_search` call.
6. Bound candidates and final results. Apply authorization before retrieval;
   treat Context filters as retrieval narrowing rather than an authorization
   boundary.
7. Preserve source keys, scores, stage provenance, and request IDs. Resolve
   source rows only through an authorized discovered surface.
8. Deduplicate by stable source identity and fit the evidence to a stated token
   budget.
9. Cite only returned evidence. State that evidence is insufficient when the
   retrieved records do not support the answer.

Polygres returns evidence. The calling agent or application composes the answer.

Select `vector_name` from the collection's registered vectors, or use its
default. For semantic plus lexical retrieval, `context_text_hybrid_search`
embeds `query` when `embedding` is omitted. On `context_joint_search`, `text`
is the semantic input and `query` supplies lexical terms.

Text queries consume retrieval allowance. Keep `use_credits` false unless the
user authorizes additional credit usage and the project has spending permission.
Keep one retry identity for the same query and use a new one for changed input.
If generation or model setup is needed, use the embedding-management tools
listed in [`mcp-tool-contract.md`](https://skills.polygres.com/polygres-sdk/references/mcp-tool-contract.md) within the user's requested scope.