InfiniteStream v2.0.0 is here — talk to your test sessions

[jonathaneoliver]

InfiniteStream v2.0.0 — talk to your test sessions

v2.0.0 is out, and the headline is two ways to interrogate a streaming session with an LLM.

💬 An in-dashboard AI chat bot

Ask plain-English questions about any session right in the browser — "why did this stall at 0:42?" — and get answers grounded in the actual analytics data (it queries the session's events, network requests, and control timeline for you). Provider-agnostic: bring your own model — Anthropic, any OpenAI-compatible endpoint, or fully local (Ollama/MLX) — and opt-in (inert until you configure a provider).

🛠️ Claude Code CLI session querying

Prefer the terminal? Point the Claude Code CLI at the rig and analyse sessions through prose, using the bundled skills — triage to survey what broke in seconds, investigate to drill a single event, forensics for "why does this keep happening." Backed by a playback-knowledge standards library and a searchable findings store.

---

Also in v2.0.0

• A new Vue 3 dashboard replacing the legacy static-HTML pages.
• Three coherent ClickHouse analytics tables sharing one severity-tagged label vocabulary that drives every chip, tint, and filter.
• A harness CLI covering the full v2 API surface (with snapshot/undo discipline).
• A player ABR characterization framework plus an Automated Testing dashboard page.
• A server-behavior control-surface test suite (rate / delay / loss / faults / transport).

⚠️ This is a major release — several v1 surfaces were removed. Read the Breaking changes before upgrading.

📦 Full release notes & images: https://github.com/jonathaneoliver/infinite-streaming/releases/tag/v2.0.0