Compress Context
compress_contextCompress a raw conversation transcript into a structured context block that preserves user goal, decisions, and state for another AI to continue.
Instructions
Compresses a raw conversation transcript into a structured ContextBridge handoff block.
Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| provider | No | Optional provider override for the extraction lane: openai, openai-chat, google, groq, custom, or hosted. | |
| transcript | Yes | The raw conversation transcript to compress. |
Output Schema
| Name | Required | Description | Default |
|---|---|---|---|
| goal | Yes | The user's primary goal as of the latest meaningful state in the transcript. | |
| version | Yes | Schema version for ContextBridge output. | |
| open_threads | Yes | Unresolved work, risks, or follow-ups that still need attention. | |
| recent_pivot | Yes | ||
| current_state | Yes | A concise description of where the work currently stands, weighted toward the final 25% of the conversation. | |
| key_artifacts | Yes | Important outputs, files, prompts, or assets mentioned in the transcript. | |
| decisions_made | Yes | Only decisions explicitly confirmed, chosen, or accepted by the user. Suggestions alone do not count. | |
| conversation_type | Yes | The dominant type of work the conversation is doing. | |
| suggested_next_step | Yes | The best immediate next action for continuing the conversation. | |
| user_expertise_level | Yes | Estimated expertise level implied by the user's language and requests. | |
| conversation_assessment | Yes | ||
| constraints_and_preferences | Yes | Important user preferences, constraints, or non-negotiables. |