Data Planning Agent

# Example Organizational Context This directory contains example context files that demonstrate how to customize the Data Planning Agent for your organization. ## Purpose Context files are markdown documents that provide the AI agent with organizational knowledge: - Company-specific terminology - Standard operating procedures - Data governance policies - Technical constraints - Communication preferences ## Usage 1. Copy this directory to your own location (local or GCS) 2. Modify the example files or create your own 3. Set `CONTEXT_DIR` environment variable to point to your context directory 4. Context will be automatically loaded when the agent starts ## File Naming Files are loaded in alphabetical order and concatenated. Use numbered prefixes to control order: - `01_organization.md` - Organizational context (loaded first) - `02_sop.md` - Standard operating procedures - `03_constraints.md` - Technical and business constraints - etc. ## Example Configuration ### Local Context ```bash # .env CONTEXT_DIR=./context ``` ### GCS Context ```bash # .env CONTEXT_DIR=gs://my-company-bucket/planning-agent-context/ ``` ## File Structure Each markdown file can contain any free-form content. The agent will receive all context before every prompt. **Recommended sections**: - Organizational background - Team structure - Communication style - Domain-specific terminology - Standard operating procedures - Data governance policies - Technical constraints - Preferred analysis patterns ## Tips - Keep context files focused and concise - Update context files as policies change - Use clear headers and formatting - Include definitions for domain-specific terms - Document any "always" or "never" rules - Specify preferred terminology ## Effects Context influences all agent interactions: - **Initial questions**: Tailored to your domain and constraints - **Follow-up questions**: Consistent with your terminology - **Data PRP generation**: Aligned with your standards and requirements Context is **not** shown to end users - it silently guides the agent's behavior.