Merwynkumar/clawblink
ClawBlink : No‑code AI agent builder from chat. Describe an agent in plain English; ClawBlink builds and runs it.
What's novel
ClawBlink : No‑code AI agent builder from chat. Describe an agent in plain English; ClawBlink builds and runs it.
Code Analysis
13 files read · 4 roundsA no-code AI agent builder that parses plain-English chat messages (via Telegram or WhatsApp) into YAML agent configs using an LLM, then runs scheduled/polling action pipelines (HTTP fetch → LLM summarize → chat notification) on a local Python runtime.
Strengths
Clean, well-organized architecture with clear separation between builder, engine, providers, and interfaces; the URL validation pipeline (intent-based fallbacks, content error detection, LLM-suggested replacements) is a thoughtful touch that makes the 'no-code' promise actually work; the SmartProvider fallback chain and channel-isolated config directories show practical engineering.
Weaknesses
Zero test coverage across 50 source files is a significant gap for a project that makes real HTTP calls and manages background threads; the scheduler loop uses a simple 60-second tick with in-memory counters rather than a proper scheduling library, which can drift and doesn't handle timezone changes or DST; the WhatsApp bridge duplicates most of the Telegram bot's command-handling logic rather than sharing a common command dispatcher.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
16
Contributors
1
Files
5488
Active weeks
2
Repository
Language
Python
Stars
1
Forks
0
License
MIT