Langfuse
Langfuse's daily totals. Model, environment, prompt, tokens, cost and scores. Prompt version, trace name and tags, latency and time to first token. Never prompts, inputs or outputs.
Access
- Kind
- A read-only key you paste
- What to give
- Public and secret key pair
- Why that one
- A Langfuse project's key pair. Valistry calls only the metrics endpoint, which returns totals — never a trace, never a prompt.
What it attributes on its own
- From the source
- day; model; environment; prompt name and version; trace name and tags; observations; input and output tokens; average latency and time to first token; what Langfuse says they cost
- Needs a rule or a trace
- the provider's invoice — Langfuse computes or ingests a cost, it does not bill you; team, workflow and agent unless a rule maps the environment; outcome — scores are a later slice
The addresses it calls
- https://cloud.langfuse.com
Valistry calls only these addresses. It blocks any other call and logs every call.
Connecting it
- Create a key pair. Open the project → Settings → API Keys → Create new API keys Open ↗
Langfuse shows a cost; your provider bills it. Connecting both Langfuse and a provider it traces counts the same call twice — connect one. Pick Langfuse to see cost by environment or prompt name. Pick the provider to match the invoice.