dependabit / detector/src / LLMProvider

Interface: LLMProvider

Defined in: packages/detector/src/llm/client.ts:148

Contract that all LLM provider implementations must satisfy.

Remarks

The only method called by the Detector is analyze. getSupportedModels, getRateLimit, and validateConfig exist for diagnostic and health-check purposes.

Use When

You need to plug in a custom or self-hosted language model as the classification backend for the detector.

Avoid When

You only need programmatic heuristics — constructing a stub provider that always returns an empty dependencies array has a small overhead but is safe.

Pitfalls

• Provider implementations must return valid JSON matching the LLMResponse shape. Returning plain text causes the detector to silently produce zero LLM-sourced results.
• Do NOT cache the analyze response across different model values — classification schemes differ between model versions.

Methods

analyze()

analyze(content, prompt): Promise<LLMResponse>;

Defined in: packages/detector/src/llm/client.ts:155

Analyze content and detect external dependencies

Parameters

Parameter	Type	Description
content	string	Text content to analyze (README, code, etc.)
prompt	string	Detection prompt template

Returns

Promise<LLMResponse>

LLM response with detected dependencies

---

getRateLimit()

getRateLimit(): Promise<RateLimitInfo>;

Defined in: packages/detector/src/llm/client.ts:165

Get current rate limit status

Returns

Promise<RateLimitInfo>

---

getSupportedModels()

getSupportedModels(): string[];

Defined in: packages/detector/src/llm/client.ts:160

Get list of supported models for this provider

Returns

string[]

---

validateConfig()

validateConfig(): boolean;

Defined in: packages/detector/src/llm/client.ts:170

Validate provider configuration

Returns

boolean