Python: Preserve Mistral prompt-cache usage details

[Ruiming Zhao (uuzzrm)]

Motivation & Context

Mistral chat-completion responses can include prompt cache hits under usage.prompt_tokens_details.cached_tokens. The Mistral adapter currently drops that field, so callers cannot observe cache reads through standard UsageDetails. This is especially visible in cost and telemetry reporting for cached prompts.

Description & Review Guide

• What are the major changes?
  • Map prompt_tokens_details.cached_tokens to both prompt/cached_tokens and cache_read_input_token_count in _parse_usage.
  • Add regression coverage for non-streaming responses and final streaming usage chunks.
• What is the impact of these changes?
  • Existing usage fields are unchanged. Mistral cache-read information is now available through the same standard field used by other providers.
• What do you want reviewers to focus on?
  • Whether the mapping and type guard match the shape of Mistral usage payloads and the conventions used by other provider clients.

Related Issue

Fixes#7589

Contribution Checklist

• The code builds clean without any errors or warnings
• All unit tests pass, and I have added new tests where possible
• The PR follows the Contribution Guidelines
• This PR is linked to an issue and there is no other open PR for this issue (see Related Issue above).
• This is not a breaking change. If it is a breaking change, add the breaking change label (or add "[BREAKING]" to the title prefix) — a workflow keeps the label and title prefix in sync automatically.

[Copilot]

Pull request overview

Preserves Mistral prompt-cache usage in standard framework usage details.

Changes:

• Maps cached prompt tokens to provider-specific and standard fields.
• Adds non-streaming and streaming regression coverage.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
python/packages/mistral/agent_framework_mistral/_chat_client.py	Parses cached prompt-token usage.
python/packages/mistral/tests/mistral/test_mistral_chat_client.py	Tests cache usage propagation.

---

💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.

[Ruiming Zhao (uuzzrm)]

@microsoft-github-policy-service agree

[Ruiming Zhao (uuzzrm)]

Following up on the strict integer validation feedback: the current head a1c1b44 contains the requested validation, and the reported repository checks are successful. Please re-review the current head when convenient.

[Ruiming Zhao (uuzzrm)]

Synced the existing branch with the latest upstream main in merge commit 17da48c.

The review fix itself is unchanged. The current head retains strict integer validation for cached_tokens (excluding bool) and regression coverage for string, float, and boolean payloads.

Local verification after the sync is environment-limited: the bundled Python has pytest but not the repository dependencies or uv, so collection stops before imports with ModuleNotFoundError for agent_framework. The prior focused CI checks for the fix were green; the new push has triggered the repository CI for the synced head.

The PR remains open and unmerged, awaiting the repository checks and maintainer review.

[github-actions]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/mistral/agent_framework_mistral
_chat_client.py	459	10	97%	270, 272–274, 281, 285, 289, 292–293, 295
TOTAL	45773	4233	90%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
9301	36 💤	0 ❌	0 🔥	2m 38s ⏱️

[Giles Odigwe (giles17)]

Automated Code Review

Reviewers: 5 | Confidence: 93% | Result: All clear

Reviewed: Correctness, Security Reliability, Test Coverage, Failure Modes, Design Approach

---

Automated review by giles17's agents

[Giles Odigwe (giles17)]

Hey Ruiming Zhao (@uuzzrm) please address failing checks on the PR

[Ruiming Zhao (uuzzrm)]

AI says: I addressed the failing checks in commit 2d72515cd. The failure was strict Pyright reporting an unknown cached_tokens type after the runtime Mapping check; the fix explicitly narrows that mapping to Mapping[str, Any] without changing the existing integer/non-bool validation.

Verified locally with:

• uv run --frozen pyright -p packages/mistral/pyproject.toml — 0 errors
• uv run --frozen pytest -q packages/mistral/tests/mistral/test_mistral_chat_client.py -m 'not integration' — 56 passed
• uv run --frozen ruff check ... and ruff format --check ... — passed

Prysai

[Ruiming Zhao (uuzzrm)]

AI says: The fix is pushed in 2d72515cd, and the local package checks pass. GitHub created the new workflow runs for this forked PR with action_required, so the Python workflows have not started yet and need maintainer approval before their results can be reported.

Prysai

[Ruiming Zhao (uuzzrm)]

Updated the branch with the current upstream main (27d82b1) in merge commit 3c705e0.

Re-ran the focused checks locally after the sync:

• uv run --frozen pyright -p packages/mistral/pyproject.toml ? 0 errors
• uv run --frozen pytest -q packages/mistral/tests/mistral/test_mistral_chat_client.py -m "not integration" ? 56 passed
• uv run --frozen ruff check on the changed Mistral files ? passed
• uv run --frozen ruff format --check on the changed Mistral files ? passed
• git diff --check ? passed

The fork workflow runs for this head are currently action_required, so the repository test workflows have not started and require maintainer approval. The PR remains open and ready for review.

AI assistance was used to prepare this update.