Generate Documentation from Prompts

TyneBase's AI can transform simple descriptions into comprehensive, well-structured documentation.

How It Works

Our AI pipeline uses a multi-stage process:

1. Prompt Analysis: Understanding your intent and requirements
2. RAG Context Retrieval: Finding relevant existing documentation
3. Content Generation: Creating structured, accurate content
4. Processing: Transforming your knowledge to md and ingesting it to the database.
5. Post-Processing: Formatting, linking, and quality checks

Writing Effective Prompts

Be Specific

❌ "Write about our API"

✅ "Create a comprehensive REST API reference for our user authentication endpoints, including request/response examples, error codes, and rate limiting details"

Provide Context

Create an onboarding guide for new backend engineers.

Context:
- We use Node.js with TypeScript
- Our API runs on Fly.io
- We use Supabase for database and auth
- Code reviews are required before merging

Include sections on:
1. Local development setup
2. Code style and conventions
3. Pull request workflow
4. Deployment process

Specify Format

Format	Best For
Article	Explanatory content, concepts
Guide	Step-by-step procedures
Runbook	Operational procedures
FAQ	Common questions
Reference	API docs, specifications

AI Provider Options

TyneBase supports three EU-compliant AI providers:

DeepSeek V3 — 1 credit per generation

• Best for: Quick drafts, routine documentation
• Strengths: Fastest, most economical
• Region: eu-west-1 via AWS Bedrock

Google Gemini 2.5 Flash — 1 credit per generation

• Best for: Research-heavy content, multimodal
• Strengths: Large context window, balanced quality
• Region: europe-west2 (London) via Vertex AI

Anthropic Claude Sonnet 4.5 — 2 credits per generation

• Best for: Nuanced writing, complex analysis
• Strengths: Detailed reasoning, best output quality
• Region: eu-central-1 via AWS Bedrock

Generation Settings

Setting	Options	Description
Tone	Professional, Casual, Technical	Writing style
Length	Brief, Standard, Comprehensive	Content depth
Examples	On/Off	Include code samples
RAG Context	On/Off	Use existing docs as reference

Document Lineage

Every AI-generated document tracks:

{
  "source": "ai_generated",
  "prompt_hash": "sha256:abc123...",
  "model": "deepseek-v3",
  "tokens_used": 2847,
  "credits_charged": 1,
  "rag_sources": ["doc-123", "doc-456"],
  "generated_at": "2026-01-10T14:30:00Z"
}

Best Practices

1. Review Before Publishing: AI content is created as drafts
2. Verify Technical Details: Especially code and configurations
3. Add Company Context: Include internal links and specifics
4. Iterate: Generate multiple versions and combine the best
5. Use RAG: Enable context retrieval for consistency