* docs(examples): add chat reasoning_effort examples

Add four examples in examples/mistral/chat/ demonstrating reasoning_effort
on mistral-medium-3-5:

- reasoning_response_shape.py: dump the raw response shape for
  reasoning_effort="high" vs "none" so users see the ThinkChunk /
  TextChunk JSON before consuming it.
- reasoning.py: single-turn call, iterate ThinkChunk and TextChunk in
  message.content.
- reasoning_with_streaming.py: handle streaming deltas where chunks
  arrive as ThinkChunk lists during thinking and as plain string
  fragments after thinking ends.
- reasoning_multi_turn.py: 3-turn math chain run with two replay
  strategies (keep vs drop ThinkChunks) and prints token usage so
  the cost difference is visible.

* address review: keep ThinkChunks for MM3.5; use timeout_ms

- reasoning_multi_turn.py: drop the keep-vs-drop comparison and
  recommend keeping ThinkChunks across turns. Per reviewer feedback,
  dropping the reasoning trace degrades MM3.5 performance.
- All four files: replace httpx.Client(timeout=...) with the SDK's
  timeout_ms parameter; remove the httpx import.

* fix lint: remove unused ThinkChunk import in reasoning_multi_turn

* simplify multi-turn: append response.choices[0].message directly

Avoids the unnecessary AssistantMessage(content=content) re-wrap and
forwards any future fields on AssistantMessage automatically.

Verified end-to-end: 3/3 runs of the math chain produce 391 -> 1173 ->
1073, and history inspection confirms each AssistantMessage slot
preserves [ThinkChunk, TextChunk].

* Workflows: Event decryption

* Improvements

* Improvements

* Decryption for event stream endpoint

* Add new route to stream decoding

* Fix a generation issue

* Do not trigger SDK generation

* ci: add required examples gate

* ci: add required lint gates

…r_provider (mistralai#517)

Add a set_tracer_provider(client, provider) helper that attaches a custom
TracerProvider to a specific Mistral client instance. Spans produced by
that client are emitted through the custom provider; other instances
continue using the global provider.

Usage:
    from mistralai.extra.observability import set_tracer_provider
    set_tracer_provider(client, my_provider)

Generated by Mistral Vibe.

Co-authored-by: Mistral Vibe <vibe@mistral.ai>

…-regen-25492313905] 2.4.5 (mistralai#518)

* ## Python SDK Changes:
* `mistral.workflows.executions.stream()`:  `response.[].data.data.union(CustomTaskInProgressResponse).attributes.payload` **Changed** (Breaking ⚠️)
* `mistral.workflows.events.get_stream_events()`:  `response.[].data.data` **Changed** (Breaking ⚠️)
* `mistral.workflows.events.get_workflow_events()`:  `response.events[]` **Changed** (Breaking ⚠️)
* `mistral.events.get_stream_events()`:  `response.[].data.data.union(CustomTaskInProgressResponse).attributes.payload` **Changed** (Breaking ⚠️)
* `mistral.events.get_workflow_events()`:  `response.events[]` **Changed** (Breaking ⚠️)
* `mistral.beta.connectors.get_authentication_methods()`:  `response.[].has_default_credentials` **Added**

* chore: align pyproject.toml and uv.lock to version 2.4.5

---------

Co-authored-by: speakeasybot <bot@speakeasyapi.dev>
Co-authored-by: GitHub Action <action@github.com>