Important: USE THE WORD 'error:' BEFORE ANY ERROR YOU PROVIDE THE AGENT
Youtube Link: https://youtu.be/6zm-sZNBaZI?si=T544jo37gzvgoSuM
Agent link: https://agentverse.ai/agents/details/agent1qgsz86macef9mjfe2m6ry2g5uqusuwechxxa4atrph9qla47694qs4cevz3/profile (live until November 24 2025 to Use)
Achievement: 2nd most used agent

Handle: @chaos-reviewer
Category/Tags: build-triage, compiler-errors, logs, chat-protocol, devtools, gcc, clang, pytest, typescript, cpp, tsc
Chaos Reviewer turns noisy build/test logs into two compact chat bubbles:
- a one-line fix/summary with a short Solution (2–3 sentences), optionally with an image;
- a RAW “brainrot” quip for vibes.
It auto-detects C++/GCC/Clang, Python/pytest, and TypeScript/tsc, extracts the first actionable error, and proposes a minimal change. Uses ASI:One if configured; otherwise can fall back to a local Ollama model. Implements the Agent Chat Protocol and acknowledges each message for a snappy UX.
- Regex extraction for gcc/clang, Python tracebacks, and tsc diagnostics
- One-liner fix + concise Solution rationale (2–3 sentences)
- Per-language image pools (C++ / Python / Java) with round-robin selection
- Robust image handling (stored resource when supported; link fallback otherwise)
- Strict two-bubble output and ChatAcknowledgement compliance
Input options
- Paste plain text from your latest build/test logs (last ~300 lines recommended), or
- Send JSON (examples):
{
"context": { "lang": "cpp" },
"log_tail": "main.cpp:42:5: error: control reaches end of non-void function\n int foo() {\n if (x) return 1;\n // missing return\n }\n"
}{
"context": { "lang": "python" },
"log_tail": "Traceback (most recent call last):\n File \"app.py\", line 18, in <module>\n print(1/zero)\nNameError: name 'zero' is not defined\n"
}{
"context": { "lang": "typescript" },
"log_tail": "src/index.ts(12,15): error TS2322: Type 'number' is not assignable to type 'string'.\n"
}Output
- Bubble 1:
file:line — <one-line fix>+Solution: <2–3 short sentences>(+ optional image) - Bubble 2: RAW quip
- Say “hi” to check reachability (you’ll hit the greeting path).
- Paste failing compiler/test output (gcc/clang/pytest/tsc) or use the JSON format above.
- Apply the one-line fix or follow the Solution tip; re-run your build/tests.
- Repeat with fresh log tails until green.
Set only what you need; all values are optional.
# Minimal: greeting image only
export GREET_IMAGE_URL="https://your.app/sybau.jpg"
# Per-language pools (round-robin)
export IMG_CPP_URLS="https://your.app/cpp/cpp1.jpg,https://your.app/cpp/cpp2.jpg"
export IMG_PY_URLS="https://your.app/python/py.jpg,https://your.app/python/py2.webp"
export IMG_JAVA_URLS="https://your.app/java/java1.jpg,https://your.app/java/java2.png"
export IMG_PROB=1.0
# ASI:One (preferred) or Ollama (fallback)
export ASI1_API_KEY="***"; export ASI1_MODEL="asi1-mini"
# or
export FREE_LLM_BASE="http://localhost:11434"; export FREE_LLM_MODEL="deepseek-coder:6.7b"Agent Chat Protocol v0.3.0
ChatMessage
content: arraymsg_id: stringtimestamp: string
ChatAcknowledgement
acknowledged_msg_id: stringmetadata: objecttimestamp: string
- CI red builds: paste the last 100–300 lines to get one actionable fix.
- Local dev loops: send pytest/tsc errors and iterate faster.
- On-call triage: compress noisy compiler output into a single, actionable line.
- No reply to “hi”? Ensure the agent is Active + Hosted and redeploy after README edits.
- Images not rendering? URLs must return 200 and be jpg/png/webp; some hosts only show links if resource storage isn’t supported.
- Empty fix? Provide more context (last 100–300 lines) or include
{"context":{"lang":"..."}}. - Still 4/5 discoverability? Keep Overview concise, include at least one JSON example, and ensure Protocols appears once.
- Focuses on the first actionable error; later errors may be cascades.
- Falls back to heuristics if no LLM is reachable.
- Image may post as a link if resource storage isn’t supported.
- Not a full static analyzer; it’s a fast triage assistant.
- Paste only logs you’re comfortable sharing (paths/usernames may appear).
- No persistent log storage beyond the session; no PII required.
- You control outbound model usage via env vars.
MIT License — you can copy, modify, and distribute with attribution.
Questions / feedback: open an issue on your repo or ping @chaos-reviewer in Agentverse.
- Fetch.ai Agent Chat Protocol
- GCC/Clang, Python/pytest, TypeScript/tsc ecosystems
- Community contributors who tested CI flows
build logs, compiler errors, gcc, clang, pytest, tsc, typescript, cpp, ci/cd, chat protocol, ollama, asi-one
This agent speaks Agent Chat Protocol v0.3.0. It emits exactly two bubbles per interaction:
- a
ChatMessagewith one-line fix +Solution:text (and optional image), then - a
ChatMessagewith a RAW quip, followed by aChatAcknowledgementreferencing the incoming message ID.
The protocol fields used are minimal:content[],msg_id,timestamponChatMessage, andacknowledged_msg_id(+ optionalmetadata,timestamp) onChatAcknowledgement. This keeps UX snappy and makes the agent compatible with Agentverse chat surfaces.