Skip to main content
POST
/
prompts
/
{prompt_name}
/
resolve
curl -X POST \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "scopes": {
      "user_id": 123,
      "project": "web-app"
    },
    "query": "user authentication issues"
  }' \
  https://api.contextbase.dev/v1/prompts/chatbot-assistant/resolve

{
  "content": "SYSTEM: You are a helpful assistant for web-app support.\n\nUSER CONTEXT: User ID 123 has previously reported login issues on 2024-01-15. Recent interactions show confusion about password reset flow. User preferences: technical explanations, quick solutions.\n\nKNOWLEDGE BASE: Authentication troubleshooting guide for web-app project shows common issues with OAuth redirects...\n\nRECENT EVENTS: System maintenance completed on auth service...",
  "metadata": {
    "token_count": 1247
  }
}

​
Overview

The prompt resolve endpoint fetches compressed, data-rich content from all contexts within a specified prompt. This combines multiple contexts in their defined order to create comprehensive prompts that give your LLM contextual awareness across different data sources without context rot.

​
Path Parameters

​
prompt_name
string
required
Name of the prompt to resolve (e.g., “chatbot-assistant”)

​
Request Body

​
scopes
object
Scoping parameters to personalize the context. These filter contexts to include only data relevant to specific users, projects, or environments.
​
query
string
Optional query string for RAG contexts. This helps filter and rank relevant content, preventing context rot by including only information semantically similar to your query.

​
Response

​
content
string
required
The compressed prompt content from all contexts combined in order, optimized for prompting your LLM with comprehensive contextual awareness.
​
metadata
object
Additional metadata about the resolved prompt.

​
Example Request

curl -X POST \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "scopes": {
      "user_id": 123,
      "project": "web-app"
    },
    "query": "user authentication issues"
  }' \
  https://api.contextbase.dev/v1/prompts/chatbot-assistant/resolve

​
Example Response

{
  "content": "SYSTEM: You are a helpful assistant for web-app support.\n\nUSER CONTEXT: User ID 123 has previously reported login issues on 2024-01-15. Recent interactions show confusion about password reset flow. User preferences: technical explanations, quick solutions.\n\nKNOWLEDGE BASE: Authentication troubleshooting guide for web-app project shows common issues with OAuth redirects...\n\nRECENT EVENTS: System maintenance completed on auth service...",
  "metadata": {
    "token_count": 1247
  }
}

​
Use Cases

  • Multi-Context AI Applications: Combine user data, knowledge bases, and system information in a single prompt
  • Comprehensive AI Agents: Give AI agents awareness across multiple data sources and contexts
  • Structured Prompt Building: Create prompts with ordered contexts for specific use cases
  • Enterprise AI Systems: Build complex prompts that include user context, company knowledge, and real-time data

​
How Prompt Resolution Works

  1. Context Retrieval: All contexts within the prompt are fetched in their defined order
  2. Individual Resolution: Each context is resolved with the provided scopes and query
  3. Content Combination: Context contents are combined in the order specified in the prompt
  4. Agentic Compression: If agentic mode is enabled, an LLM further compresses the combined content to be relevant to the query
  5. Token Limiting: Result is trimmed to respect the prompt’s token limit
  6. Structured Output: Final compressed prompt ready for your LLM with metadata

​
Difference from Context Resolve

  • Context Resolve: Returns content from a single context
  • Prompt Resolve: Returns combined content from multiple contexts in a specific order
  • Use Prompts: When you need to combine different types of context (user data + knowledge base + system info)
  • Use Contexts: When you only need content from a single source or context type