ops(llm): probe OpenRouter's thinking-off shape for DeepSeek/Qwen
Maintainers usually reply within 1 day
Nobody has claimed this yet.
Assessment
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Newbie friendliness
- 42/100
- Issue type
- Bug
- Clarity
- Mostly clear
- Activity status
- Active
- Tech stack
- python
- Domain
- backend-api-design
Research direction
Read adapters/openai.py and providers/openrouter.py to compare the reasoning-off values. After #596 merges and the OpenRouter key has balance, call each model on the OpenRouter lane at temperature 0 and record the HTTP status, reasoning and completion token counts, and whether content is non-empty. Done means the results are recorded, any rejected settings are trimmed, and OpenRouter uses the canonical values.
Written by the indexing model from the issue text.
Description
Blocked until #596 merges and the OpenRouter key is topped up (#523).
#596 pins DeepSeek V4 Pro and Qwen3.7 Plus to PINNED_NO_THINKING. On CommonStack that sends extra_body.thinking = {type: "disabled"}, verified by the #539 probe. On failover to OpenRouter, adapters/openai.py sends:
{"reasoning": {"effort": "none", "enabled": false, "exclude": true}}
That combination has never been sent to OpenRouter for these models. The lane is quota-dead today, so its first real use will come during a CommonStack outage, the worst moment to find a 400 or a reply that still reasons to the cap.
Task: once the key has balance, make one call per model (DeepSeek V4 Pro, Qwen3.7 Plus) on the OpenRouter lane with the shape above at temperature 0. Record HTTP status, reasoning_tokens, completion_tokens and whether content is non-empty. If OpenRouter rejects the combination, trim it to the subset it accepts.
Related: providers/openrouter.py keeps its own _OFF_VALUES, a copy of adapters/openai.py::REASONING_OFF_VALUES. Point it at the canonical set while you're there.
🤖 Generated with Claude Code
- Dominant language
- Python
- Stars
- 767
- Forks
- 156
- Avg merge
- 22h 35m
- Merged PRs (30d)
- 81
Getting set up
- Ships a Dockerfile or Docker Compose file
- No pull request template
- No contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from Open-Finance-Lab/AgenticTrading
-
Difficulty 2/5 1-3 hours Newbie friendliness 80/100
Open-Finance-Lab/AgenticTrading#663 ·
Maintainers usually reply within 1 day
-
fix(backtests): an LLM "sell 10" sells the whole positionPossibly taken @subidhkhanal claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 84/100
Open-Finance-Lab/AgenticTrading#661 ·
Maintainers usually reply within 1 day
-
fix(llm): a non-ASCII execution handoff raises UnicodeEncodeError instead of ExecutionHandoffErrorPossibly taken @subidhkhanal claimed this 4 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 85/100
Open-Finance-Lab/AgenticTrading#659 ·
Maintainers usually reply within 1 day
-
fix(backtests): an LLM action with trailing whitespace ("buy ") is silently droppedPossibly taken @subidhkhanal claimed this 4 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 83/100
Open-Finance-Lab/AgenticTrading#658 ·
Maintainers usually reply within 1 day
-
fix(llm): Gemini adapter drops thinking tokens from the output countPossibly taken @subidhkhanal claimed this 4 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 88/100
Open-Finance-Lab/AgenticTrading#657 ·
Maintainers usually reply within 1 day
All issues in Open-Finance-Lab/AgenticTrading
Similar issues
-
follow-up
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 67/100
-
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 84/100
modelscope/modelscope#1833 ·
Maintainers usually reply within 1 day
-
ocr_data_to_elements() raises AttributeError when a layout element is a ListPossibly taken @Eapp1e claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 79/100
Unstructured-IO/unstructured#4616 ·
Maintainers usually reply within 1 day