Moda vs Langfuse

1. Difference 1Moda finds the failures. Langfuse makes you go looking.Moda's detections run on every trace from the first one, and failures are grouped into problems ranked by users affected. In Langfuse you write or pick each evaluator, choose a sampling rate, and find the failure categories by reading traces.
2. Difference 2Moda judges the whole agent run. Langfuse judges one step at a time.Agents fail across turns and tool calls, and Moda reads each run end to end. Langfuse's LLM-as-a-judge evaluators score one observation at a time and cannot be applied to sessions.

What moves depends on how your app sends traces to Langfuse today. Spans from OpenTelemetry instrumentation move by changing where they are exported. Calls traced with Langfuse's own SDK move to the Moda SDK. In both cases, carry your session and user IDs over so each agent run stays one trace under the right user, then import the history you already have in Langfuse.

1. 1.Point your OpenTelemetry exporter at Moda

   If your spans come from OpenTelemetry GenAI instrumentation, OpenLLMetry, or the Vercel AI SDK's telemetry and go to Langfuse's /api/public/otel endpoint, Moda reads them as they are. Change the exporter's endpoint to https://moda-ingest.modas.workers.dev/v1/traces and its header to Authorization: Bearer with your Moda API key. For direct export, set moda.user_id on each span to the same user ID you sent to Langfuse; changing the endpoint does not translate langfuse.user.id or user.id. If you use a Collector, step 3 copies that attribute for you.

   Moda ingest accepts uncompressed request bodies only, and the OpenTelemetry spec leaves the default compression to each SDK, so set OTEL_EXPORTER_OTLP_TRACES_COMPRESSION=none. A Collector compresses by default; the config in step 3 sets compression: none on the Moda exporter.

   Exporter environment variablesCopy

   export OTEL_EXPORTER_OTLP_TRACES_ENDPOINT="https://moda-ingest.modas.workers.dev/v1/traces"
   export OTEL_EXPORTER_OTLP_TRACES_HEADERS="Authorization=Bearer YOUR_MODA_API_KEY"
   export OTEL_EXPORTER_OTLP_TRACES_COMPRESSION="none"

2. 2.Replace Langfuse SDK tracing with the Moda SDK

   Langfuse's own SDK, including the observe decorator and the Langfuse OpenAI wrapper, writes model input and output to langfuse.observation.input and langfuse.observation.output. Moda does not read those attributes, so repointing that exporter is not enough.

   Replace those wrappers with the Moda SDK for Python or Node.js. One init call instruments the OpenAI and Anthropic clients, including streamed responses and tool calls, without changes to your provider code.

3. 3.Carry your session and user IDs over

   Moda groups spans into one trace by moda.conversation_id, or gen_ai.conversation.id, across OpenTelemetry traces, and attributes them to a user by moda.user_id. It does not read Langfuse's keys, langfuse.session.id and session.id for sessions or langfuse.user.id and user.id for users. Set moda.conversation_id to the value you used as the Langfuse session ID and moda.user_id to the Langfuse user ID; with the Moda SDK, that is moda.conversation_id and moda.user_id in Python. In Node.js, import withContext from moda-ai and wrap each request in withContext(sessionId, userId, async () => { /* your existing agent code */ }), using your Langfuse IDs. This isolates concurrent requests. Reserve the process-wide Moda.conversationId and Moda.userId setters for scripts handling one session at a time.

   If your spans pass through an OpenTelemetry Collector, the config below copies the Langfuse keys into moda.conversation_id and moda.user_id, so existing session and user IDs carry over without a code change. The copies work span by span: a session stays one trace only when every span carries the session ID, and a span is attributed to a user only when it carries the user ID. The insert action keeps any moda.user_id a span already has. Because session.id and user.id are also generic OpenTelemetry keys, drop those rules if other services send them through the same Collector with a broader meaning.

   otel-collector-config.yamlCopy

   receivers:
     otlp:
       protocols:
         grpc: {}
         http: {}
   
   processors:
     batch: {}
     attributes/moda-session:
       actions:
         - key: moda.conversation_id
           from_attribute: langfuse.session.id
           action: insert
         # Langfuse also accepts session.id. It is a generic OpenTelemetry key:
         # drop this rule if other services send session.id through this
         # Collector with a broader meaning, such as a browser session.
         - key: moda.conversation_id
           from_attribute: session.id
           action: insert
         - key: moda.user_id
           from_attribute: langfuse.user.id
           action: insert
         # Langfuse also accepts user.id, another generic key: drop this rule
         # if other services send user.id through this Collector.
         - key: moda.user_id
           from_attribute: user.id
           action: insert
   
   exporters:
     # Replaces your Langfuse exporter.
     otlphttp/moda:
       traces_endpoint: https://moda-ingest.modas.workers.dev/v1/traces
       headers:
         Authorization: Bearer ${env:MODA_API_KEY}
       # Moda ingest rejects compressed request bodies; the Collector gzips by default.
       compression: none
   
   service:
     pipelines:
       traces:
         receivers: [otlp]
         processors: [attributes/moda-session, batch]
         exporters: [otlphttp/moda]

4. 4.Import your Langfuse history

   Moda imports the traces you already have in Langfuse. In Settings, open Integrations, then Import traces, and choose Langfuse. Paste a Langfuse public key and secret key, pick your Langfuse Cloud region or enter your self-hosted URL, then choose the project and a date range.

   Each Langfuse trace arrives with its observations, inputs and outputs, models, token usage, errors, and user ID. Traces that share a Langfuse session ID become one Moda trace, and Langfuse scores, tags, and metadata are kept with the trace they belong to. The import runs in the background, and you can pause, resume, or cancel it. Prompts, datasets, and experiments stay in Langfuse.