dimies

Integrate Dimies — conversation analytics for AI agents — into any codebase in one command.

npx dimies init --key dm_live_…     # scaffold a tracking helper (Node or Python)
npx dimies test --key dm_live_…     # send a test conversation, verify end-to-end
npx dimies prompt                    # print an integration prompt for AI coding agents
npx dimies spec                      # print the machine-readable spec (llms.txt)

The fastest path: let your coding agent do it

npx dimies prompt | pbcopy

Paste into Claude Code (or Cursor, etc.) inside the repo you want instrumented. The agent fetches llms.txt from your Dimies instance and wires everything: fire-and-forget conversation tracking, trace spans around LLM/tool calls, env vars in .env.example.

What init does

• Detects Node (TS/JS) or Python
• Writes a zero-dependency dimies.ts / dimies.js / dimies.py helper exporting trackConversation(...) / track_conversation(...) — fire-and-forget, never throws, never blocks your response path
• Adds DIMIES_API_KEY / DIMIES_URL to .env.example (never hardcodes your key)

Flags

| Flag | Meaning | |---|---| | --key dm_live_… | Dimies API key (or env DIMIES_API_KEY) | | --url https://… | Your Dimies instance URL (or env DIMIES_URL) | | --dir path | Target directory for init (default .) | | --lang node\|python | Override language detection |

Tracing

Pass a trace array of spans to trackConversation(...) to get a Langfuse-style execution tree on each conversation. Give every span a stable id, and set input/output to the payload — for an llm span, the prompt and the completion; for a tool span, the args and the result:

trackConversation({
  conversation_id: chat.id,
  messages,
  trace: [
    { id: "llm-1", name: "vera.llm", type: "llm",
      model: "gpt-4o", tokens_in: 1240, tokens_out: 85,
      cost: 0.0017,            // optional — exact spend; omit and Dimies estimates it
      input: prompt, output: completion },
  ],
});

Field names are exact — prompt/completion/response are silently dropped, so a span sent that way shows "No payload recorded". Merge is field-level and keyed by id: resend the same id to update a span (e.g. { id: "llm-1", status: "error" }) and only the fields you include change — the rest is preserved.

Cost accuracy: set model, tokens_in, tokens_out on llm spans and the Costs dashboard prices them at list rates. For exact spend (negotiated rates, caching), also send cost in USD from your provider's usage — it's used verbatim instead of the estimate.

Full span schema: <your-instance>/llms.txt.

MIT.