This week we launched the most comprehensive view of AI tool adoption yet—going beyond downloads to track real-world usage across GitHub, Bitbucket, and 8 major developer marketplaces.
🚀 New This Week: Dependency Tracking
We now track actual installations in production codebases, not just downloads. Here's what we found scanning 61,000+ repositories:
NPM Real-World Adoption (Top 5)
- openai: 40,192 repos
- langchain: 9,152 repos
- ollama: 8,256 repos
- @ai-sdk/openai: 7,040 repos
- anthropic: 5,104 repos
Python Real-World Adoption (Top 5)
- openai: 21,408 repos
- chromadb: 16,112 repos
- langchain-openai: 13,952 repos
- ollama: 13,456 repos
- anthropic: 9,936 repos
Key insight: While OpenAI leads in raw numbers, Ollama's presence in both NPM (8K) and Python (13K) repos shows local LLMs are gaining serious traction in production environments.
📊 Multi-Provider Setups Emerging
We detected 12 repos using both openai + anthropic in their package.json, and 8 repos with the same combo in requirements.txt. This represents a new pattern: developers hedging against single-provider dependency.
Co-occurrence patterns we're tracking:
- openai + anthropic: 12 repos (NPM), 8 repos (Python)
- langchain + ollama: 8 repos (NPM), 11 repos (Python)
- @ai-sdk/anthropic + @ai-sdk/openai: 9 repos
This is the start of the "multi-model era" where applications switch between providers based on task, cost, or availability.
🏪 Marketplace Expansion
We now track AI tools across 8 developer marketplaces:
VS Code Marketplace
- GitHub Copilot: 55.2M installs
- Claude Code: 1.2M installs
- Continue: 756K installs
- Cursor extension: 89K installs
JetBrains Marketplace (Surprise Leader)
- JetBrains AI Assistant: 57.7M downloads
- GitHub Copilot: 31.8M downloads
- Tabnine: 5.4M downloads
Finding: JetBrains AI Assistant beats Copilot by 81% in the JetBrains ecosystem. This suggests professional developers on paid IDEs have different adoption patterns than VS Code users.
🔍 Historical Tracking Now Live
We're now backfilling commit history to track when dependencies were added:
- Analyzing commits back to June 2022 (start of AI coding tools era)
- Tracking first adoption dates per tool
- Identifying early adopters and laggards
- Building growth curves from real git history
Why this matters: Download counts lie. Package installs in dead repos lie. Git commits showing active use? That's the signal.
📈 What We're Watching
1. Ollama's Rise
From 0 to 21K total repos (NPM + Python) in 28 months. Compare to:
- Docker Hub: 34.5M pulls
- GitHub repos: 21K
- Cross-platform: Strong in both ecosystems
Prediction: Ollama hits 50K repos by Q1 2026.
2. Chromadb Dominance
16,112 Python repos use chromadb for vector storage. This is higher than the actual LLM libraries in many cases.
Insight: The infrastructure layer (vector DBs, embeddings) is growing faster than the model layer. This suggests maturing architectures.
3. Multi-Provider Patterns
Every week we see more repos with both Anthropic + OpenAI SDKs. This is the hedging behavior we predicted in September.
Next: We're building a "Provider Switching Detector" to track when repos migrate between models.
🛠️ Data Methodology
All data collected October 10-17, 2025 using:
- GitHub Code Search API (authenticated, rate-limited)
- Official marketplace APIs where available
- Web scraping with 3-5s delays (respectful)
- PostgreSQL for time-series storage
- Historical backfill from git commit history
Total data points this week:
- 61,000+ GitHub repos scanned
- 8 marketplaces tracked
- 14 NPM packages monitored
- 14 Python packages monitored
- 50+ high-star repos analyzed for historical patterns
💡 What This Means
For developers: Your choice of AI tool is no longer isolated. Multi-provider setups are becoming standard for production applications. Plan for provider switching from day one.
For tool makers: Marketplace presence matters less than real installations. We found tools with 50M+ marketplace "installs" but only 5K in actual package.json files.
For investors: Look past vanity metrics. The tools appearing in requirements.txt and package.json files of funded startups—that's your signal.
🔮 Coming Next Week
- Bitbucket dependency tracking (enterprise adoption patterns)
- Historical backfill completion (June 2022 → present)
- Provider switching detection algorithm
- Cross-marketplace correlation analysis
📊 Access the Data
All findings backed by real data:
- GitHub adoption: 61K+ repos scanned
- Marketplace presence: 8 platforms tracked
- Historical trends: June 2022 → October 2025
Get the full dataset: vibe-data.com
Data collected October 10-17, 2025. Methodology: GitHub Code Search API, marketplace APIs, git history analysis. All metrics verified against multiple sources.
Follow @vibe_data for real-time updates on AI development trends.