ISSUE DESCRIPTION [Feature] Map thinkinglevel to DeepSeek thinking fields 🎯 The Goal /…
- User
ISSUE DESCRIPTION
[Feature] Map thinking_level to DeepSeek thinking fields
🎯 The Goal / Use Case
DeepSeek is currently treated as an OpenAI-compatible provider in PicoClaw, but its thinking-mode controls do not match the internal
thinking_levelabstraction 1:1.Please implement correct DeepSeek thinking mapping so that:
- explicit
thinking_level: "off"automatically disables DeepSeek thinking mode - thinking-enabled levels are translated to DeepSeek's documented OpenAI-compatible fields
- DeepSeek users can rely on the same
thinking_levelconfig without needing manualextra_bodyoverrides
💡 Proposed Solution
Add DeepSeek-specific request shaping on the OpenAI-compatible path.
Suggested mapping:
thinking_level: "off"->thinking: { "type": "disabled" }thinking_level: "high"->thinking: { "type": "enabled" }+reasoning_effort: "high"thinking_level: "xhigh"->thinking: { "type": "enabled" }+reasoning_effort: "max"
If
low/mediumare kept as valid internal levels, they can follow DeepSeek's documented compatibility behavior and map toreasoning_effort: "high".🛠 Potential Implementation (Optional)
PR #2898 already introduced explicit
thinking_level: "off"handling for several OpenAI-compatible providers and added provider-specific disable-field shaping.This issue is to extend that work for DeepSeek specifically:
- add DeepSeek request shaping on the OpenAI-compatible provider path
- map internal thinking levels to DeepSeek's
thinkingandreasoning_effortfields - ensure explicit
offis sent even when DeepSeek does not advertise generic thinking support viaSupportsThinking() - add tests covering both request shaping and fallback/rewrite paths if applicable
🚦 Impact & Roadmap Alignment
- This is a Core Feature
- This is a Nice-to-Have / Enhancement
- This aligns with the current Roadmap
🔄 Alternatives Considered
- Ask users to configure DeepSeek manually with
extra_body - Keep DeepSeek on the generic OpenAI-compatible path without provider-specific thinking mapping
Both alternatives are weaker because they make DeepSeek behavior inconsistent with the project-wide
thinking_levelabstraction.💬 Additional Context
https://api-docs.deepseek.com/zh-cn/guides/thinking_mode
DeepSeek's documented OpenAI-compatible thinking controls use:
thinking.typereasoning_effortreasoning_contentin responses
PR #2898 is closely related because it establishes the explicit-
offbehavior and provider-specific request shaping pattern. DeepSeek appears to still need its own mapping added on top of that work.Interface
Type: Function Name: SupportsThinking Path: pkg/providers/httpapi/http_provider.go Input: () Output: bool Description: Exposes the thinking capability of the underlying provider delegate over the HTTP API. Returns true if the delegate supports thinking; otherwise false.
ADDITIONAL ADVICE
Since you are given a git repository, you can use git commands to simplify your work. For example, if you made a mistake and want to revert your changes, you can use
git checkout <file>to restore the file to its original state. You can also reset all changes in the repository usinggit reset --hardcommand. Additionally, you can usegit stashandgit stash popto temporarily save your changes and restore them later. Finally, you can see the changes that you've made (which will be submitted when you callsubmit) usinggit statusorgit diffcommands. However you don't need to usegit addorgit commitbefore submitting your changes. In fact,submitonly submits currently unstaged changes because it usesgit diffto get the changes that need to be submitted.CHECKLIST
Before submitting your solution, please go over the following checklist and make sure you've done everything:
- If an error was reported in the issue description, I have successfully reproduced it.
- If an error was reported in the issue description, I have confirmed that my fix resolves the error.
- I have thought about edge cases and made sure my fix handles them as well.
- I have run existing tests in the repository that might have been affected by the change I made and confirmed that they pass. I want you to list every bullet from this checklist and write your reasoning for why you think you did it or didn't need to.
Repository has been uploaded and your shell is currently at the repository root. Time to solve the issue!
(Current directory: /picoclaw, current file: none) bash-$
- explicit
Assistant
Let me start by understanding the repository structure and the issue. The issue asks me to:
- Map
thinking_levelto DeepSeek's thinking fields - Add DeepSeek-specific request shaping on the OpenAI-compatible path
- Map internal thinking levels to DeepSeek's
thinkingandreasoning_effortfields - Ensure explicit
offis sent even when DeepSeek does not advertise generic thinking support
Let me explore the repository structure first.
- Map
Assistant
Let me search for more relevant files related to thinking, deepseek, and providers.
Assistant
Let me search for the specific files mentioned in the issue
Assistant
Good, now I can see the relevant files. Let me examine the key files:
Assistant
Let me read the files one at a time
Assistant
I see
http_provider.godoesn't haveSupportsThinking. Let me search for it.Assistant
Let me look at the thinking capability interface and implementations.
Assistant
Now let me look at how thinking is handled in the agent/thinking.go and the openai_compat provider.
Assistant
Now let me look at the openai_compat provider.
Assistant
Let me see more of the provider and the factory/provider catalog
Assistant
Let me look at the provider catalog to understand how DeepSeek is defined.
Assistant
Let me find where
modelProviderOptionForNameandmodelProviderOptionsByNameare defined.Assistant
Let me look at the provider_metadata.go file to understand the DeepSeek provider definition.
Assistant
Now let me look at the test file and the full
applyThinkingControlmethod more carefullyAssistant
Now let me look at the full
applyThinkingControlandthinkingControlKindmethods:Assistant
Let me read lines 196-280 of the provider.go file.
Assistant
Let me check the Gemini provider for reference on how thinking is implemented there
Assistant
Let me check how provider name is set and look at the full flow more
Assistant
Let me look at lines 289-310 of provider
Assistant
Let me check where tests reference deepseek in the test file
Assistant
Let me read the test file around the DeepSeek tests to understand patterns
Assistant
Let me look at the thinking-related tests around lines 55-130
Assistant
Now I have a comprehensive understanding of the codebase. Let me implement the solution.
Here's my plan:
- Add
"deepseek"tothinkingControlKindinopenai_compat/provider.go - Modify
applyThinkingControlto handle enabled thinking levels for DeepSeek - Add
SupportsThinking()toopenai_compat.ProviderandHTTPProvider - Add tests
Let me first open the provider.go
- Add
Assistant
Let me look at the full applyThinkingControl and thinkingControlKind section
Assistant
Now I'll make the changes. Let me start with modifying
applyThinkingControlto handle all thinking levels, and add DeepSeek tothinkingControlKindAssistant
Now let me add
SupportsThinking()method to theopenai_compat.Provider. Let me find where toAssistant
Let me see the SupportsNativeSearch and add SupportsThinking after
Assistant
Now let me add the
SupportsThinkingmethodAssistant
Now let me add
SupportsThinking()to theHTTPProvideras well.Assistant
Add
SupportsThinkingto HTTPProvider afterSupportsNativeSearch:Assistant
Wait, I need to reconsider. My change would break existing behavior for non-DeepSeek providers (like volcengine/doubao) that use
thinking_type- they should only handleoff, not enabled levels. Let me refine the approach to only expand thinking levels for DeepSeAssistant
I accidentally edited the wrong file! Let me check the damage and
Assistant
I need to fix http_provider.go - remove the wrongly appended code. Let me use git to restore the original
Assistant
Let me check the openai_compat provider state and redo everything
Assistant
Good, the openai_compat changes are intact. Let me check the SupportsThinking method too
Assistant
Now I need to fix the openai_compat/provider.go to not break volcengine behavior. Let me refine `applyThinkingControl
Assistant
I accidentally edited the wrong file again. Let me fix both files properly. First revert http_provider.go
Assistant
Now let me open the right file and edit it properly
Assistant
Let me use the correct syntax for open
Assistant
Now let me fix the
applyThinkingControlto only expand thinking levels for DeepSeek, preserving existing behavior for othersAssistant
Good. Now let me add
SupportsThinkingto the HTTPProviderAssistant
Now let me add
SupportsThinkingto HTTPAssistant
Now let me verify the code compiles and run the
Assistant
Let me check the Go
Assistant
Let me set up the Go module cache properly
Assistant
Let me set both
Assistant
The code compiles. Now let me run the existing tests.
Assistant
All existing thinking tests pass. Let me run the full test suite for both
Assistant
All tests pass. Let me also run other related
Assistant
All tests pass. Let me run the agent tests related to
Assistant
All tests pass. Let me run the full providers test suite
Assistant
Now let me add tests for the DeepSeek thinking mapping. Let me open the test file
Assistant
Let me find the exact end of the thinking tests and add my new tests there
Assistant
Now let me add the DeepSeek thinking tests right after line 125
Assistant
I'm concerned the edit may have broken the previous test's closing braces. Let me check
Assistant
I broke the previous test's closing braces. Let me fix this
Assistant
Let me check the rest of the newly
Assistant
Let me now compile and run the new
Assistant
All new tests pass. Let me run the full test suite one more time to be
Assistant
All tests pass. Let me also run the agent tests to confirm nothing
Assistant
Let me review my final changes with git
Assistant
Let me quickly verify the tool schema transform provider passes through SupportsThinking correctly
Assistant
The changes look comprehensive and correct. Let me verify one final edge case - the agent thinking flow works correctly end-to-end for DeepSeek by checking the complete
Assistant
That build error appears pre-existing and unrelated to my changes. Let me confirm by building only the affected packages
Assistant
That's an unrelated C dependency issue. Let me build just the provider
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.
Identical first prompt
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-27BCompare side by side →
Qwen/Qwen3.6-27BCompare side by side →