Alida Vector

Product knowledge changes constantly. Alida keeps your chatbot's knowledge base current while catching noisy, missing, or unexpected content before users see it.

Alida Vector crawls websites, optionally with a headless browser for JavaScript-heavy pages; ingests Jira Service Management documentation and files from cloud storage or disk; verifies the result; and stores it in pgvector.

View on GitHub Read the docs

02:00  scheduled refresh started

sources   1,284 checked
diff          7 changed, 2 removed
review    LLM verifier passed
human     no review needed
live      vector DB updated

Automated pipeline

Gather

Pull fresh content from websites, rendered pages, files, and object storage.

Compare

Build a candidate index and diff it against the current live run.

Validate

Use deterministic checks and an LLM verifier to review the changes.

Activate

Passes go live automatically. Suspicious runs wait for inspection.

Configuration

One scheduled job

Configure sources, validation, and activation once. Let the job keep the retrieval index fresh.

Deployment docs

schedule: "0 2 * * *"

sources:
  - type: website
    sitemap_url: https://example.com/sitemap.xml
  - type: s3
    bucket: product-docs

validation:
  verifier: openai
  activate_when: pass

Alida Vector is made by Filip de Waard. Questions? Feedback? Feature requests? You can reach me at fmw@vix.io.