feat(providers): Add Cerebras Reasoning Support - #10417

Closed

aaalaniz wants to merge 45 commits into

aaif-goose:mainfrom

aaalaniz:phase2-reasoning-refactor

Commits on Jul 21, 2026

feat: refactor Reasoning to enum and add ReasoningConfig
aaalaniz
committed
92e1282
feat: add ThinkingPreservationFormat and fix provider reasoning tests
aaalaniz
committed
34c2e26
feat(providers): Refactor ReasoningConfig to support advanced configuration
aaalaniz
committed
8f8e5ff
fix(providers): Resolve clippy warnings in openai tests
aaalaniz
committed
ecc10cc
chore: Fix formatting
aaalaniz
committed
7803834
fix(providers): Address Codex PR feedback
aaalaniz
committed
6a77aca
fix anthropic_def test to use ModelInfo::new instead of Default
aaalaniz
committed
78f07bd
style: cargo fmt
aaalaniz
committed
4469a5f
Propagate declarative ReasoningConfig into requests
aaalaniz
committed
8cde23f
chore: cargo fmt
aaalaniz
committed
9dc5f52
fix: safely propagate declarative reasoning configs
aaalaniz
committed
6460426
fix: reasoning format and extra body handling
aaalaniz
committed
6c693ef
fix: safely merge stream_options with include_usage
aaalaniz
committed
63615a5
style: cargo fmt
aaalaniz
committed
34d0dc1
fix: surface declarative reasoning config in model info
aaalaniz
committed
e2cc30b
fix: preserve reasoning heuristics for default custom models
aaalaniz
committed
06c8fd7
fix(providers): preserve explicit reasoning=false for custom models
aaalaniz
committed
cec7b89
fix(cli): handle Option<Reasoning> in configure
aaalaniz
committed
4962fbe
fix(sdk): map Option<bool> to Option<Reasoning> in bindings
aaalaniz
committed
b81b100
fix(providers): preserve reasoning config for custom ollama models
aaalaniz
committed
f19fffa
fix(providers): prefer custom reasoning settings over canonical defaults in stream
aaalaniz
committed
bb1d55d
fix(providers): preserve explicit reasoning disable for custom models
aaalaniz
committed
68b533d
Fix explicit reasoning override being discarded
aaalaniz
committed
569a481
chore: format and fixes
aaalaniz
committed
2d50bde
fix(providers): update tests to use Reasoning enum
aaalaniz
committed
eaed111
fix(providers): address latest PR feedback on reasoning logic
aaalaniz
committed
7d1e9a2
fix(providers): resolve clippy warning on unnecessary lazy evaluation
aaalaniz
committed
0b8a6e7
feat(cerebras): configure reasoning for gpt-oss-120b and zai-glm-4.7
aaalaniz
committed
91be969
feat(cerebras): map reasoning_effort to thinking levels
aaalaniz
committed
d091c0e
Propagate declarative reasoning config to ModelConfig in agent.rs
aaalaniz
committed
df69695
Apply suggestion from @aaalaniz
aaalaniz
committed
7b943e9
Fix reasoning provider fallback for custom models that omit reasoning metadata
aaalaniz
committed
00e300e
Honor 'Off' for custom reasoning effort by omitting reasoning parameters for non-OpenAI models
aaalaniz
committed
9053f3f
Add effort_mapping to cerebras declarative models to allow Off to map to none
aaalaniz
committed
aba3abc
Simplify cerebras effort_mapping
aaalaniz
committed
5695b86
Simplify zai-glm-4.7 effort_mapping
aaalaniz
committed
6812b2f
Add gemma4
aaalaniz
committed
b3754b8
Fix reasoning_is_explicit persistence and remove ToSchema
aaalaniz
committed
8f8f4ee
Apply Ollama thinking preservation formats
aaalaniz
committed
a70c1e9
Re-add thinking_preservation_format to gemma-4-31b
aaalaniz
committed
7535b75
Set preserves_thinking: true for Cerebras provider
aaalaniz
committed
7fff0a2
Merge reasoning config in update_provider
aaalaniz
committed
5c8c4db
Merge provider reasoning for non-explicit defaults
aaalaniz
committed
975c5fc
Let per-model preservation enable thinking replay
aaalaniz
committed
4798767

Commits on Jul 27, 2026

Merge origin/main into phase2-reasoning-refactor

aaalaniz
and
claude
committed
81d9fbb