yash91sharma/MALTopic-py
A multi-agent LLM topic modeling library
What's novel
A multi-agent LLM topic modeling library
Code Analysis
10 files read · 3 roundsA Python library that uses sequential LLM API calls (via OpenAI) to enrich survey free-text responses with structured data, extract topics from the enriched text, and optionally deduplicate overlapping topics, with a Streamlit GUI wrapper.
Strengths
Clean separation of concerns across core, LLM client, prompts, utils, and stats modules; comprehensive test suite with good mocking of LLM interactions covering edge cases like token limits, invalid JSON, and fallback paths; practical batching logic with tiktoken-based token counting and graceful degradation.
Weaknesses
The 'multi-agent' framing is misleading — it's really three sequential LLM calls with different prompts, not independent agents; the core logic is primarily prompt engineering and API orchestration rather than novel algorithms, and the consolidation step is explicitly a naive name-based dedup; only OpenAI is supported despite the README mentioning Gemini and Llama as future options.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
24
Contributors
1
Files
19
Active weeks
7
Repository
Language
Python
Stars
3
Forks
0
License
MIT