/**
* Phase 3 - Tier 3 Documentation Topics
*
* 7 specialized topics:
* - recipe-rag
* - agent-patterns
* - guide-testing
* - guide-versioning
* - providers-openai
* - providers-anthropic
* - providers-google
*/
export const DOCS_RECIPE_RAG = `# RAG Patterns
> Retrieval Augmented Generation
## Overview
RAG combines retrieval with generation for knowledge-grounded responses.
## Basic RAG Pattern
\\\`\\\`\\\`
<system>
You are a helpful assistant. Use the following context to answer questions.
If the context doesn't contain the answer, say so.
Context:
{{ context }}
</system>
<user>{{ question }}</user>
\\\`\\\`\\\`
## With Built-in Search
\\\`\\\`\\\`yaml
tools:
- latitude/search
\\\`\\\`\\\`
\\\`\\\`\\\`
Search for relevant information about {{ topic }} and answer the question.
Question: {{ question }}
\\\`\\\`\\\`
## Document Chunks
\\\`\\\`\\\`
<system>
Answer based on these documents:
{{ for doc in documents }}
[Document {{ loop.index }}]
{{ doc.content }}
---
{{ endfor }}
</system>
<user>{{ question }}</user>
\\\`\\\`\\\`
## Tips
- Keep context focused
- Cite sources when possible
- Handle missing information gracefully
## Next Steps
- \\\`latitude_get_docs({ topic: "tools-builtin" })\\\`
- \\\`latitude_get_docs({ topic: "conversation-history" })\\\`
`;
export const DOCS_AGENT_PATTERNS = `# Agent Collaboration Patterns
> Multi-agent workflows and orchestration
## Overview
Complex tasks benefit from specialized agents working together.
## Orchestrator Pattern
\\\`\\\`\\\`yaml
type: agent
agents:
- agents/planner
- agents/executor
- agents/reviewer
\\\`\\\`\\\`
\\\`\\\`\\\`
Coordinate to complete: {{ task }}
1. Planner creates the plan
2. Executor implements each step
3. Reviewer validates results
\\\`\\\`\\\`
## Specialist Pattern
\\\`\\\`\\\`yaml
type: agent
agents:
- agents/researcher
- agents/writer
- agents/editor
\\\`\\\`\\\`
## Debate Pattern
\\\`\\\`\\\`yaml
agents:
- agents/proponent
- agents/opponent
- agents/judge
\\\`\\\`\\\`
## Handoff Patterns
Sequential: A → B → C
Parallel: A + B → C
Conditional: if X then A else B
## Best Practices
| Practice | Description |
|----------|-------------|
| Clear roles | Each agent specialized |
| Defined interfaces | What each passes |
| Error handling | When agents fail |
## Next Steps
- \\\`latitude_get_docs({ topic: "agents" })\\\`
- \\\`latitude_get_docs({ topic: "chains" })\\\`
`;
export const DOCS_GUIDE_TESTING = `# Testing Prompts
> Validate and evaluate prompts
## Testing Strategies
1. **Manual testing**: Run with sample inputs
2. **Edge case testing**: Unusual inputs
3. **Regression testing**: After changes
4. **A/B testing**: Compare versions
## Test Workflow
\\\`\\\`\\\`
1. Create draft version
2. Push prompt
3. Run with test inputs
4. Evaluate outputs
5. Iterate or publish
\\\`\\\`\\\`
## Test Cases
| Test Type | Example |
|-----------|---------|
| Happy path | Normal expected input |
| Edge case | Empty, very long, special chars |
| Error case | Invalid format, missing data |
| Boundary | Min/max values |
## Evaluation Criteria
- Accuracy: Is output correct?
- Consistency: Same input → similar output
- Format: Follows schema?
- Safety: No harmful content?
## Using Logs
\\\`\\\`\\\`
latitude_run_prompt → get response
latitude_create_log → record for analysis
\\\`\\\`\\\`
## Next Steps
- \\\`latitude_get_docs({ topic: "guide-versioning" })\\\`
- \\\`latitude_get_docs({ topic: "guide-debugging" })\\\`
`;
export const DOCS_GUIDE_VERSIONING = `# Version Management
> Draft vs live workflows
## Version Types
| Type | Description |
|------|-------------|
| Draft | Work in progress, not deployed |
| Live | Production version |
## Workflow
\\\`\\\`\\\`
1. latitude_create_version → new draft
2. latitude_push_prompt → add/update prompts
3. Test thoroughly
4. Publish to live
\\\`\\\`\\\`
## Best Practices
- Always test in draft first
- Use descriptive version names
- Document changes
- Keep backups of working versions
## Managing Versions
\\\`\\\`\\\`
latitude_list_versions → see all versions
latitude_get_prompt → inspect specific prompt
\\\`\\\`\\\`
## Rollback Strategy
If issues in live:
1. Create new draft from previous good version
2. Fix issues
3. Test
4. Publish
## Next Steps
- \\\`latitude_get_docs({ topic: "guide-testing" })\\\`
- \\\`latitude_list_versions\\\`
`;
export const DOCS_PROVIDERS_OPENAI = `# OpenAI Provider
> GPT-4 and GPT-3.5 specifics
## Models
| Model | Best For |
|-------|----------|
| gpt-4o | Complex tasks, multimodal |
| gpt-4o-mini | Fast, cost-effective |
| gpt-3.5-turbo | Simple tasks, high volume |
## Configuration
\\\`\\\`\\\`yaml
---
provider: OpenAI
model: gpt-4o
temperature: 0.7
---
\\\`\\\`\\\`
## Features
- Function calling (tools)
- JSON mode
- Vision (images)
- Streaming
## JSON Mode
\\\`\\\`\\\`yaml
---
provider: OpenAI
model: gpt-4o
schema:
type: object
properties:
answer:
type: string
---
\\\`\\\`\\\`
## Best Practices
- Use gpt-4o for quality
- Use gpt-4o-mini for speed/cost
- Temperature 0 for factual
- Include clear system prompts
## Next Steps
- \\\`latitude_get_docs({ topic: "config-basics" })\\\`
- \\\`latitude_get_docs({ topic: "config-generation" })\\\`
`;
export const DOCS_PROVIDERS_ANTHROPIC = `# Anthropic Provider
> Claude specifics
## Models
| Model | Best For |
|-------|----------|
| claude-3-opus | Most capable |
| claude-3-sonnet | Balanced |
| claude-3-haiku | Fast, efficient |
## Configuration
\\\`\\\`\\\`yaml
---
provider: Anthropic
model: claude-3-sonnet
temperature: 0.7
---
\\\`\\\`\\\`
## Features
- Long context (200k tokens)
- Vision (images)
- Function calling
- Constitutional AI
## Strengths
- Following instructions precisely
- Long document analysis
- Nuanced reasoning
- Safety-focused
## Best Practices
- Use system prompts for behavior
- Leverage long context for RAG
- Clear, structured instructions
- Use for complex analysis
## Next Steps
- \\\`latitude_get_docs({ topic: "config-basics" })\\\`
- \\\`latitude_get_docs({ topic: "recipe-rag" })\\\`
`;
export const DOCS_PROVIDERS_GOOGLE = `# Google Provider
> Gemini specifics
## Models
| Model | Best For |
|-------|----------|
| gemini-pro | General purpose |
| gemini-ultra | Most capable |
| gemini-flash | Fast responses |
## Configuration
\\\`\\\`\\\`yaml
---
provider: Google
model: gemini-pro
temperature: 0.7
---
\\\`\\\`\\\`
## Features
- Multimodal (text, images, video)
- Long context
- Function calling
- Grounding with Google Search
## Strengths
- Multimodal understanding
- Google Search integration
- Efficient processing
- Wide language support
## Best Practices
- Use for multimodal tasks
- Leverage search grounding
- Clear structured prompts
## Next Steps
- \\\`latitude_get_docs({ topic: "config-basics" })\\\`
- \\\`latitude_get_docs({ topic: "messages-multimodal" })\\\`
`;
