How tulasinayak/depscan works
depscan scans Python repositories for dependencies and vulnerabilities, locates vulnerable-package usage, then optionally assesses impact as affected, not affected, or needs review. The principal flow is scan → optional repository context → vulnerability check, exposed through CLI and Streamlit interfaces. Checks use deterministic code analysis and, where needed, an LLM; repository code is parsed, never executed. The README documents evaluation and suite workflows as additional product capabilities. Some subsystem internals are unsampled, so their uncertain connections are omitted.
Main entry points: orchestrator.py Orchestrator · client.py LLM client · cli.py CLI · stepwise.py Stepwise check
Components 24
User interfaces
- CLI
depscan/cli.py - Streamlit app
app.py - Check page
depscan/ui/check_page.py - UI components
depscan/ui/components.py
Repository scanning
- Repository mapper
depscan/agents/repo_mapper.py - Manifest parser
depscan/parsers/python_manifests.py - CVE matcher
depscan/agents/cve_matcher.py - Usage locator
depscan/agents/usage_locator.py - Repository context
depscan/agents/repo_context.py
Impact analysis
- Orchestrator
depscan/orchestrator.py - Reachability index
depscan/codeindex.py - Trigger specs
depscan/triggers.py - Stepwise check
depscan/agents/stepwise.py - Holistic check
depscan/agents/exploitability.py - Package grounding
depscan/grounding.py
External services and state
- OSV.dev
- LLM client
depscan/llm/client.py - Local LLM
- Cache
depscan/cache.py
Results and evaluation
- Saved results
- Report rendering
depscan/report.py - Evaluation
depscan/evaluate.py - Evaluation suites
depscan/suite.py
Other components
- Engineer
Connections 30
- Engineer → CLI: runs commands
README.md - Engineer → Streamlit app: opens interface
README.md - Streamlit app → Check page: routes to
app.py - Check page → UI components: renders via
depscan/ui/check_page.py - CLI → Orchestrator: invokes scans and checks
depscan/cli.py - UI components → Orchestrator: invokes analysis
depscan/ui/components.py - Orchestrator → Repository mapper: maps repository
depscan/orchestrator.py - Repository mapper → Manifest parser: parses manifests
depscan/agents/repo_mapper.py - Orchestrator → CVE matcher: matches advisories
depscan/orchestrator.py - CVE matcher → OSV.dev: queries advisories
README.md - Orchestrator → Usage locator: locates usage
depscan/orchestrator.py - Orchestrator → Repository context: builds context
depscan/orchestrator.py - Orchestrator → Stepwise check: runs six gates
depscan/orchestrator.py - Stepwise check → Reachability index: checks reachability
depscan/agents/stepwise.py - Orchestrator → Reachability index: constructs index
depscan/orchestrator.py - Stepwise check → Trigger specs: uses trigger specs
depscan/agents/stepwise.py - Orchestrator → Trigger specs: loads trigger specs
depscan/orchestrator.py - Orchestrator → Holistic check: runs holistic check
depscan/orchestrator.py - Orchestrator → Package grounding: uses package facts
depscan/orchestrator.py - Stepwise check → LLM client: asks narrow questions
README.md - Holistic check → LLM client: requests verdict
depscan/agents/exploitability.py - Repository context → LLM client: requests summary
depscan/agents/repo_context.py - Trigger specs → LLM client: generates specs
depscan/triggers.py - LLM client → Local LLM: sends prompts
depscan/llm/client.py - CVE matcher → Cache: caches advisories
depscan/agents/cve_matcher.py - Orchestrator → Saved results: saves scan and verdicts
depscan/orchestrator.py - CLI → Report rendering: shows results
depscan/cli.py - CLI → Evaluation: evaluates results
depscan/cli.py - Evaluation suites → Evaluation: scores cases
depscan/suite.py - Evaluation suites → Orchestrator: runs suite analyses
depscan/suite.py