User Tools

Site Tools


doc:llmusage



How to use LLM within your IDE

Documentation | Other Specifications | LLMs Usage

Specialized AI index for crawlers (llms.txt) is available, Google and other crawlers supporting this feature should be aware soon,

Complete, AI-optimized API documentation for integrating TatukGIS with language models, code generation, and intelligent applications.

Available Resources by Platform

Platform api_index.jsonl llms.txt
Master Index - llms.txt
Delphi jsonl llms.txt
Python jsonl llms.txt
.NET/C# jsonl llms.txt
Java jsonl llms.txt
ActiveX/COM jsonl llms.txt
Combined jsonl -

Choose Your Approach

Different workflows, different tools. Here's what works best:

Use Case Recommended Approach Why
Real-time API search MCP Server Direct integration, all 5 platforms, instant lookup
Quick API lookup Claude web UI Instant, no setup required
Code generation in IDE GitHub Copilot Real-time, IDE-native
Terminal/automation Claude Code CLI Scriptable, CLI-friendly
Privacy-critical work Ollama/LM Studio No cloud, offline, local only
Team projects Claude Projects Shareable, persistent context
Production RAG pipeline api_index.jsonl + Vector DB Scalable, semantic search

The MCP (Model Context Protocol) server provides a superior alternative to JSONL files, offering real-time API search, structured queries, and seamless integration with Claude tools. It's the recommended approach for dynamic API lookups.

MCP Server URL: https://docs.tatukgis.com/DK/api/mcp

Claude Desktop

Connect to TatukGIS DK directly in Claude Desktop for seamless API lookups while writing code.

Setup:

1. Open Claude Desktop Settings → Connectors 2. Click Add New Connector 3. Enter:

  1. Name: TatukGIS DK

4. Restart Claude Desktop

Usage:

  1. Ask: “Search TatukGIS DK for Layer methods”
  2. Or: “Show me how to use Viewer in Python”
  3. Claude will search across all 5 platforms automatically

Claude Code (CLI)

Connect the TatukGIS DK MCP server to Claude Code for direct API access in your terminal.

Setup:

Add to your Claude config (typically ~/.claude/claude.json on Mac/Linux or %APPDATA%\Claude\claude.json on Windows):

{
  "mcpServers": {
    "tatukgis-dk": {
      "url": "https://docs.tatukgis.com/DK/api/mcp",
      "description": "TatukGIS DK - All platforms (Delphi, Python, .NET, Java, ActiveX)"
    }
  }
}

Restart Claude Code - The server will be available as a tool in conversations.

Usage:

  1. Search TatukGIS DK for “Viewer” class across all platforms
  2. Ask for code examples in any supported language

Alternative: Using JSONL + LLM Tools

If MCP is not available in your workflow, use JSONL files directly with these tools.

Claude (Web UI)

Best for: Quick API questions, code generation, learning

Setup:

1. Go to [[https://claude.ai|claude.ai]]
2. Click the attachment icon (📎)
3. Upload ''api_index.jsonl'' from your platform
4. Ask your question

Example prompts:

  • “How do I create a viewer component?”
  • “Generate code to load a shapefile”
  • “What methods are available on TGIS_Layer?”

Claude Code (CLI with JSONL)

Best for: Developers, terminal users, automation

Setup:

Create an alias:

<code bash>

alias delphi-api='claude –file https://docs.tatukgis.com/api/delphi/api_index.jsonl' delphi-api “how do I render a map?”

</code>

Ollama (Local LLM)

Best for: Privacy, offline use, no API costs

Setup:

1. Install from [[https://ollama.ai|ollama.ai]]
2. Download a model: ''ollama pull mistral''
3. Save your ''api_index.jsonl'' locally

Usage:

<code bash>

# Query the local LLM with API context curl http://localhost:11434/api/generate -d '{

"model": "mistral",
"prompt": "Here is the API reference: [paste api_index.jsonl contents]\n\nQ: How do I create a layer?",
"stream": false

}'

</code>

Popular models: Mistral 7B (fast), Llama 2 (strong), Neural Chat (Q&A optimized)

LM Studio (GUI)

Best for: Non-technical users, visual interface

Setup:

1. Download from [[https://lmstudio.ai|lmstudio.ai]]
2. Load a model (e.g., Mistral)
3. Open chat window

Usage:

1. Copy ''api_index.jsonl'' contents
2. Paste into LM Studio chat
3. Ask your question
4. Get responses without cloud upload

FAQs

Q: Which tool is fastest? A: Claude (web) and Copilot (IDE) are fastest. Ollama/LM Studio are slower but free and private.

Q: Can I use my own API key? A: Yes. Claude Code and all tools accept API keys. Local LLMs (Ollama) need no API key.

Q: Is my code/data safe? A: Local LLMs (Ollama, LM Studio) keep everything offline. Cloud models (Claude, ChatGPT) follow their privacy policies.

Q: Can I embed this in my app? A: Yes. Use api_index.jsonl in your RAG pipeline (Pinecone, Weaviate, ChromaDB) for production.

Q: What's the cost? A: Free LLMs (Ollama, LM Studio) = $0. ChatGPT = per-token. Claude = per-token. GitHub Copilot = $10/month.

Remarks

Other LLM and tools (e.g., Embarcadero KAI) supporting MCP should also benefit from connecting to TatukGIS MCP server.

2026/09/28 12:43

Page Tools