LCORE-1560: Output vector stores in user-facing format

[asimurka]

Description

Currently, LCORE users specify vector store IDs in user-facing format that is internally translated into LLS format. POST v1/responses endpoint echoes configured tools in tools attribute that currently contains vector store IDs in LLS format.
This PR introduces reverse translation to user-facing format during response processing.

Type of change

• Refactor
• New feature
• Bug fix
• CVE fix
• Optimization
• Documentation Update
• Configuration Update
• Bump-up service version
• Bump-up dependent library
• Bump-up library or tool used for development (does not change the final image)
• CI configuration change
• Konflux configuration change
• Unit tests improvement
• Integration tests improvement
• End to end tests improvement
• Benchmarks improvement

Tools used to create PR

Identify any AI code assistants used in this PR (for transparency and review context)

• Assisted-by: Cursor

Related Tickets & Documents

• Related Issue #
• Closes # LCORE-1560

Checklist before requesting a review

• I have performed a self-review of my code.
• PR has passed all pre-merge test jobs.
• If it is a core feature, I have added thorough tests.

Testing

• Please provide detailed steps to perform tests related to this code change.
• How were the fix/results from this change verified? Please provide relevant screenshots or results.

Summary by CodeRabbit

• New Features

  • Vector store IDs are translated automatically in requests and responses so user-facing IDs are accepted and returned consistently for file search tools.

• Documentation

  • Clarified that file search accepts LCORE-format IDs and that the system performs internal translation to/from the internal format; added a dedicated "Vector store IDs" note explaining request/response behavior.

[coderabbitai]

Walkthrough

Adds LCORE↔Llama Stack vector-store-ID translation: new utility to remap file_search.vector_store_ids, integrated into request echoing and both streaming and non‑streaming response generation; docs updated and unit tests added. (46 words)

Changes

Cohort / File(s)	Summary
Documentation docs/responses.md	Clarified that file_search.vector_store_ids are accepted/returned in LCORE format and translated internally to/from Llama Stack format.
Translation Utility src/utils/tool_formatter.py	Added translate_vector_store_ids_to_user_facing(tools, rag_id_mapping) to remap file_search vector_store_ids using a provided mapping; preserves unmapped IDs and non-file_search tools.
Request Echoing src/models/requests.py	ResponsesRequest.echoed_params() now applies translate_vector_store_ids_to_user_facing(..., configuration.rag_id_mapping) when emitting the tools field for synthetic/OpenAI-style echoed params.
Response Handling src/app/endpoints/responses.py	During streaming and non‑streaming response assembly, response tools are rewritten via translate_vector_store_ids_to_user_facing(...) using configuration.rag_id_mapping before sending to clients.
Tests tests/unit/utils/test_tool_formatter.py	New unit tests covering empty mapping, mapped/unmapped IDs, non-file_search tool preservation, input immutability for file_search entries, and mixed-list ordering/behavior.

Sequence Diagram

sequenceDiagram
    participant Client
    participant RequestHandler as Request Handler
    participant Translator as ID Translator
    participant CoreLogic as Core Logic
    participant ResponseGen as Response Generator

    Client->>RequestHandler: Send request with LCORE `vector_store_ids`
    RequestHandler->>Translator: echoed_params(tools, rag_id_mapping)
    Translator->>RequestHandler: Return tools with Llama Stack IDs
    RequestHandler->>CoreLogic: Invoke core logic using Llama Stack IDs
    CoreLogic->>ResponseGen: Produce response containing tools (Llama Stack IDs)
    ResponseGen->>Translator: translate_vector_store_ids_to_user_facing(tools, rag_id_mapping)
    Translator->>ResponseGen: Return tools with LCORE IDs
    ResponseGen->>Client: Stream/return response with LCORE `vector_store_ids`

Loading

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
Title check	✅ Passed	The title 'LCORE-1560: Output vector stores in user-facing format' directly summarizes the main change—translating vector store IDs to user-facing format in response outputs, which is the core objective across all modified files.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 3

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@src/app/endpoints/responses.py`:
- Around line 486-493: The response currently only translates top-level
response["tools"] but misses translating response["tool_choice"] when
tool_choice.type == "allowed_tools", leaving internal vector DB IDs in the
echoed payload; update the code that handles chunk_dict["response"] (and the
similar block at the other response path) to call
translate_vector_store_ids_to_user_facing for both response["tools"] and
response["tool_choice"]["tools"] (when tool_choice is a dict and type ==
"allowed_tools"), using configuration.rag_id_mapping and preserving other
tool_choice fields (i.e., replace tool_choice.tools with the translated list
while keeping the rest of tool_choice intact) so both streaming and
non-streaming responses return user-facing IDs.

In `@src/models/requests.py`:
- Around line 768-780: The response currently translates data["tools"] but not
data["tool_choice"], so internal llama-stack vector IDs can be echoed; after you
assign data["tools"] using translate_vector_store_ids_to_user_facing, also
translate the echoed tool choice(s) using the same helper and mapping: locate
the code around model_dump / tools handling (symbols: model_dump, tools,
OutputToolMCP, translate_vector_store_ids_to_user_facing,
configuration.rag_id_mapping) and, if data contains "tool_choice", call
translate_vector_store_ids_to_user_facing on that value (or map its contained
IDs) to replace internal vector_store IDs with user-facing IDs before returning
data.

In `@src/utils/tool_formatter.py`:
- Around line 151-152: The function currently returns the original tools list
when rag_id_mapping is falsy, violating the "return new list" contract; change
the early return to return a shallow copy of tools (e.g., tools.copy() or
list(tools)) so the caller receives a new list object while preserving the same
tool dict identities—modify the return at the rag_id_mapping falsy check in
tool_formatter (the block containing "if not rag_id.mapping: return tools")
accordingly.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 893fc8d4-2cd3-4fda-9d54-0258a71049a8

📥 Commits

Reviewing files that changed from the base of the PR and between 39116c7 and c03bec1.

📒 Files selected for processing (5)

• docs/responses.md
• src/app/endpoints/responses.py
• src/models/requests.py
• src/utils/tool_formatter.py
• tests/unit/utils/test_tool_formatter.py

[tisnik]

LGTM

[are-ces]

LGTM