safedep/vet
1
0
Fork 0
mirror of https://github.com/safedep/vet.git synced 2025-12-10 13:43:01 -06:00
Code Issues Packages Projects Releases 68 Wiki Activity
vet/docs/agent.md
Abhisek Datta 5f4cccbc85
feat: Add Support for Agentic Query and Analysis (#535) 
* Add initial UI for agent mode

* fix: Cleanup and define agent contract

* Add react agent

* Add interactions memory

* Add support for stdio based MCP integration

* Add basic sqlite3 report generator

* fix: Persist vulnerabilities with package relation

* fix: Persist license information

* refactor: Agents into its own command package

* feat: Add support for tool calling introspection

* refactor: UI to hide implementation detail

* sqlite3 reporter persist dependency graph

* fix: Support multiple LLM provider for agent

* docs: Update agents doc

* docs: Remove deprecated query docs

* fix: UI tests

* fix: Linter issue

* Add support for prompt mode

* Improve UI with animation

* Fix UI tests after update

* Add OpenSSF scorecard persistence

* Add slsa provenances in sqlite3 reporter

* Add test cases for sqlite3 reporter

* Fix agent doc

* fix: Sqlite3 reporter use safe accessors

* feat: Add support for fast model

* feat: Simplify and streamline agent UI for better user experience

- Remove decorative borders and excessive styling to maximize output area
- Implement clean minimal design similar to modern TUI interfaces
- Add bordered input area for clear visual separation
- Move thinking indicator above input area for better visibility
- Enhance input field reset logic for proper line alignment
- Remove verbose help text and status messages
- Optimize layout calculations for full width utilization
- Add smooth animations for agent thinking state with spinner
- Clean up code structure and remove unused progress bar functionality

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* fix: Improve agent status line

* test: Update UI tests

* fix: Use terminal safe rendering

* fix: Fix nil deref without storing empty strings in DB

* fix: Support overwriting sqlite3 database

* fix: Data model to use m2m between manifest and package

* style: Fix linter issue with unused variables

* Misc fixes

* Add test for agent memory

---------

Co-authored-by: Claude <noreply@anthropic.com>
2025-07-11 18:37:44 +05:30

1.2 KiB
Raw Blame History

Agents

vet natively supports AI agents with MCP based integration for tools.

To get started, set an API key for the LLM you want to use. Example:

export OPENAI_API_KEY=sk-...
export ANTHROPIC_API_KEY=sk-...
export GEMINI_API_KEY=AIza...

Note: You can also set the model to use with OPENAI_MODEL_OVERRIDE, ANTHROPIC_MODEL_OVERRIDE and GEMINI_MODEL_OVERRIDE environment variables to override the default model used by the agent.

Fast Mode

All agents support a --fast flag to use a faster LLM model instead of a slower but more powerful reasoning model. This is only for influencing the default choice of model. It can be overridden by setting model provider specific environment variables such as OPENAI_MODEL_OVERRIDE, ANTHROPIC_MODEL_OVERRIDE and GEMINI_MODEL_OVERRIDE.

Available Agents

Query Agent

The query agent helps run query and analysis over vet's sqlite3 reporting database. To use it:

  • Run a vet scan and generate report in sqlite3 format
vet scan --insights-v2 -M package-lock.json --report-sqlite3 report.db

Note: Agents only work with --insights-v2

  • Start the query agent
vet agent query --db report.db
  • Thats it! Start asking questions about the scan results.