bug: langchain-tracer and openai-tracer packages use Socket.IO against a backend that has no Socket.IO server #41

Description

The monorepo still contains and versions two tracer packages that cannot work with the current backend:

• packages/langchain-tracer (@axon-ai/langchain-tracer) — src/client.ts connects via Socket.IO to http://localhost:8000 by default.
• packages/openai-tracer (@axon-ai/openai-tracer) — src/OpenAITracer.ts connects via Socket.IO to http://localhost:3000 by default.

The current backend is OTLP-native: it exposes POST /v1/traces (OTLP JSON/protobuf) on port 4000 and pushes dashboard updates over SSE. There is no Socket.IO server anywhere in the backend, and neither default port matches. Anyone who installs these packages gets endless connection retries and no traces, no matter how they configure the endpoint.

They also drag socket.io-client into the workspace dependency tree for nothing.

Suggested fix

Pick one per package:

1. Rewrite them as thin OTel instrumentations that emit spans to the standard OTLP endpoint (the backend already understands OpenLLMetry/OpenInference-style attributes), or
2. Deprecate and remove them from the workspace, and mark the npm packages deprecated with a pointer to the OTel setup snippets the dashboard already shows on the Overview screen.

Given the README onboarding is already OTel-first, option 2 is likely the lower-cost path — keeping broken packages published under the @axon-ai scope hurts trust more than having fewer packages.