GitLab

The GitLab Datasource connects a GitLab repository to Cognipeer as a searchable knowledge source. Cognipeer recursively fetches all files from the specified project and branch, indexes their content, and makes it available for your Peer to retrieve during conversations.

---

Use Cases

• Enable developers to ask questions about a codebase hosted on GitLab.
• Index Markdown-based documentation living in a GitLab repository.
• Power engineering chatbots or internal knowledge portals.
• Support code review and onboarding by letting Peers explain repository structure.

---

Prerequisites

• A GitLab Personal Access Token with read_api and read_repository scopes.
• The numeric Project ID of the GitLab repository (visible on the project's main page).
• The branch or tag you want to index (default: main).

---

Setting Up a GitLab Datasource

1. Navigate to Datasources and click Add Datasource.
2. Select GitLab as the datasource type.
3. Fill in the required fields:
   • Project ID: The numeric ID of the GitLab project (e.g., 12345678).
   • Access Token: Your GitLab Personal Access Token.
   • Reference (optional): Branch or tag to index (defaults to main).
4. Click Save.
5. Click Sync to begin indexing.

---

What Gets Indexed

Cognipeer fetches the full repository tree and processes all text-based files (up to 5,000 files). File types indexed include source code, documentation (.md, .rst, .txt), and configuration files (.json, .yaml, .toml). Binary files are skipped.

---

Keeping the Datasource Up to Date

GitLab datasources do not auto-sync when new commits are pushed. To update:

• Click Sync manually in the datasource settings.
• For automated sync workflows from GitLab CI, use the Developer Hub.

---

Best Practices

• Use descriptive commit messages and file names to improve search result quality.
• For large monorepos, consider splitting into multiple datasources scoped to specific paths.
• Rotate access tokens periodically and update the datasource credentials accordingly.

---

Limitations

• Maximum 5,000 files per sync.
• Only text-based files are indexed; binary and compiled assets are skipped.
• Pagination is handled automatically, but very large repositories may reach the file limit.