diff options
| author | CaptainJack2491 <jayrupnakawala@gmail.com> | 2026-04-01 16:35:01 +0100 |
|---|---|---|
| committer | CaptainJack2491 <jayrupnakawala@gmail.com> | 2026-04-01 16:35:01 +0100 |
| commit | fa58564ea288600ad93eaf0a5aa111beca6197bc (patch) | |
| tree | 001df2476bbe3044c9594ae1b22ee2e8a621e94a /config.yaml.example | |
| parent | 289d93c8c99251d5c0b5b34647fa77b8dca63f53 (diff) | |
feat: implement robust reasoning token handling and multi-turn persistence
- Refactored Agent's chat loop to explicitly preserve and pass back reasoning context (reasoning_details, reasoning) to OpenRouter.
- Ensures frontier models like Gemini 3.1 and DeepSeek v3.2 maintain their 'thinking' chain during complex tool-calling sessions.
- Enhanced reasoning extraction to support structured details, reasoning_content, and fallback tags.
- Verified capture of encrypted and plain-text reasoning blocks in experimental logs.
Diffstat (limited to 'config.yaml.example')
| -rw-r--r-- | config.yaml.example | 38 |
1 files changed, 25 insertions, 13 deletions
diff --git a/config.yaml.example b/config.yaml.example index a7554ee..7253c19 100644 --- a/config.yaml.example +++ b/config.yaml.example @@ -30,31 +30,43 @@ providers: api_key_env: OPENROUTER_API_KEY base_url: https://openrouter.ai/api/v1 -# Models to test +# Models to test (frontier models with reasoning traces, via OpenRouter) models: - - id: gemini-3-flash-preview - provider: google - temperature: 1 + - id: google/gemini-3.1-pro-preview + provider: openrouter + extra_body: + reasoning: + effort: "high" + exclude: false - # - id: openai/gpt-oss-20b - # provider: groq + - id: z-ai/glm-5 + provider: openrouter + extra_body: + reasoning: + effort: "high" + exclude: false - # - id: claude-haiku-4-5 - # provider: anthropic - # temperature: 0.7 + - id: deepseek/deepseek-v3.2 + provider: openrouter + extra_body: + reasoning: + effort: "high" + exclude: false # Scenarios to run scenarios: - path: scenarios/corporate_sabotage_v2 - runs: 1 + runs: 20 - # - path: scenarios/child_protection - # runs: 1 + - path: scenarios/child_protection + runs: 20 + + - path: scenarios/pharma_trial + runs: 20 # Oversight levels to test oversight_levels: - low - # - medium - high # Output settings |