pd.log()

Log the result of an LLM call back to VersionR. This is how results appear in the dashboard — latency trends, token usage, and output history all come from pd.log().

Logging is fire-and-forget — it never throws or delays your application.

Signature

pd.log(entry: LogEntry): Promise<void>

Parameters

Parameter	Type	Required	Description
`entry.promptId`	`string`	Yes	The `promptId` from `pd.get()` or `pd.render()`
`entry.promptSlug`	`string`	Yes	The `promptSlug` from the prompt response
`entry.versionId`	`string`	Yes	The `versionId` from the prompt response
`entry.versionNumber`	`number`	Yes	The `versionNumber` from the prompt response
`entry.environment`	`string`	Yes	The environment used (e.g. `'production'`)
`entry.renderedPrompt`	`string`	Yes	The final prompt text sent to the LLM
`entry.output`	`string`	Yes	The LLM’s response text
`entry.latencyMs`	`number`	Yes	Time from request sent to response received (ms)
`entry.tokens.prompt`	`number`	No	Prompt tokens used
`entry.tokens.completion`	`number`	No	Completion tokens used
`entry.score`	`number`	No	Optional quality score (0–1) for evals
`entry.metadata`	`Record<string, string>`	No	Arbitrary key/value pairs attached to this log entry

Example

const startTime = Date.now()
 
const rendered = await pd.render('interviewer-system', {
  env: 'production',
  variables: {
    candidate_name: 'Eoin',
    job_role: 'Senior Engineer',
    difficulty: 'hard',
  },
})
 
const response = await openai.chat.completions.create({
  model: rendered.model,
  temperature: rendered.temperature,
  messages: [
    { role: rendered.role, content: rendered.content },
    { role: 'user', content: userMessage },
  ],
})
 
await pd.log({
  promptId: rendered.promptId,
  promptSlug: rendered.promptSlug,
  versionId: rendered.versionId,
  versionNumber: rendered.versionNumber,
  environment: 'production',
  renderedPrompt: rendered.rendered,
  output: response.choices[0]?.message.content ?? '',
  latencyMs: Date.now() - startTime,
  tokens: {
    prompt: response.usage?.prompt_tokens ?? 0,
    completion: response.usage?.completion_tokens ?? 0,
  },
})

With metadata

Attach arbitrary metadata to a log entry for filtering in the dashboard:

await pd.log({
  promptId: rendered.promptId,
  promptSlug: rendered.promptSlug,
  versionId: rendered.versionId,
  versionNumber: rendered.versionNumber,
  environment: 'production',
  renderedPrompt: rendered.rendered,
  output: responseText,
  latencyMs: elapsed,
  metadata: {
    userId: 'user_abc123',
    sessionId: 'sess_xyz',
    region: 'eu-west-1',
  },
})

Notes

pd.log() resolves immediately — the network request runs in the background. Your app does not wait for it.
If the log request fails (network error, service outage), the failure is silently swallowed. It will not throw or affect your application.
Cost estimation (visible in the dashboard) is computed server-side from token counts + model name. Provide tokens.prompt and tokens.completion for accurate cost tracking.

Log entries are stored in ClickHouse and queryable from the dashboard. High-volume applications can safely call pd.log() on every request.

pd.render()Error Handling & Caching