Repository setup, parser configuration, and index tuning.
Core settings are configured via environment variables. These are defined in env/dev.env for development or env/prod.env for production.
| Variable | Required | Description |
|---|---|---|
| DATABASE_URL | Yes | PostgreSQL connection string for Auth/Contract DB |
| REDIS_URL | Yes | Redis connection string for sessions and rate limiting |
| DVN_SERVICE_KEY | Yes | Internal service-to-service auth token (min 32 chars) |
| DVN_RE_BEARER_TOKEN | Yes | Token for DVN-ReServices access (same as DVN_SERVICE_KEY) |
| DVN_CONDUCTOR_BEARER_TOKEN | Yes | Dedicated token for Conductor AI service |
| DVN_GPU_HOST | No | Set to true on GPU-equipped hosts for MinerU PDF processing |
# Minimum required env/dev.env
DATABASE_URL=postgresql://dvn:password@postgres:5432/dvn_auth
REDIS_PASSWORD=<strong-password>
REDIS_URL=redis://:password@redis:6379/0
DVN_SERVICE_KEY=<random-32-chars>
Each schema can have multiple sources. Supported source types:
| Type | Configuration | Best For |
|---|---|---|
| Git | URL, branch, access token (or SSH key) | Active development, CI/CD integration |
| Local | Filesystem path | On-premise, static archives |
| ZIP | Upload via UI | One-time analysis, quick evaluation |
| SharePoint | OAuth-delegated site/drive access | Documents already managed in Microsoft 365 |
| WebDAV | Server URL, credentials | Network shares and on-prem document servers |
Parsers are auto-selected based on file extension and content analysis. You can override the default parser per file extension in the Schema Manager.
Available parsers include:
On mainframe platforms, file extensions are often missing or unreliable. Codegraphs supports logical extension overrides — you can map any file to any parser without renaming the file on disk.
# Example: Map member "ACCT001" to COBOL parser
Schema → Files → ACCT001 → Set Logical Extension → cbl
This only affects indexing behavior. The original file is never modified.
| Setting | Default | Description |
|---|---|---|
| Exclude patterns | — | Gitignore-style patterns for files to skip, via a .noindex file |
| Max file size | 10 MB | Skip files larger than this threshold |
| Incremental indexing | on | Only re-index changed files on subsequent runs |
| Comment indexing | on | Comment lines are indexed separately and searchable via the comment: operator |