OpenAI-Compatible Agents

Any agent that can make HTTP requests can use Synapse Layer. This guide shows the REST API pattern for adding encrypted persistent memory to OpenAI-compatible agents, custom LLM wrappers, or any HTTP-capable system.

Overview

Not every agent framework has a dedicated Synapse Layer integration. For GPT-based agents, custom LLM wrappers, or any system that can make HTTP calls, the REST API provides full access to the Cognitive Security Pipeline — PII sanitization, intent classification, differential privacy, and AES-256-GCM encryption.

Note

For Python agents, consider using the Python SDK instead — it wraps the REST API with error handling, retries, and type safety.

Store Memory

Store a memory via the REST API. The content is automatically processed through the full security pipeline before persistence:

bash
curl -X POST https://your-forge-url/api/v1/memory/commit \
  -H "Content-Type: application/json" \
  -H "x-connect-token: YOUR_TOKEN" \
  -d '{
    "content": "User prefers dark mode and TypeScript.",
    "agentId": "my-agent",
    "metadata": {
      "source": "onboarding",
      "importance": 0.8
    }
  }'

Python equivalent using httpx:

python
import httpx

response = httpx.post(
    "https://your-forge-url/api/v1/memory/commit",
    headers={
        "Content-Type": "application/json",
        "x-connect-token": "YOUR_TOKEN",
    },
    json={
        "content": "User prefers dark mode and TypeScript.",
        "agentId": "my-agent",
    },
)
result = response.json()
print(f"Memory ID: {result['memoryId']}")
print(f"Trust Quotient: {result['trustQuotient']}")

Recall Memory

bash
curl -X POST https://your-forge-url/api/v1/sdk/recall \
  -H "Content-Type: application/json" \
  -H "x-connect-token: YOUR_TOKEN" \
  -d '{
    "query": "user preferences",
    "agentId": "my-agent",
    "topK": 5
  }'

The response includes matched memories with content, trust quotient scores, and metadata for each result.

Function Calling Pattern

For OpenAI function-calling agents, define Synapse Layer as a tool:

python
tools = [
    {
        "type": "function",
        "function": {
            "name": "store_memory",
            "description": "Store information for later recall. Content is encrypted and PII-sanitized.",
            "parameters": {
                "type": "object",
                "properties": {
                    "content": {
                        "type": "string",
                        "description": "The information to remember"
                    }
                },
                "required": ["content"]
            }
        }
    },
    {
        "type": "function",
        "function": {
            "name": "recall_memory",
            "description": "Search stored memories by semantic similarity.",
            "parameters": {
                "type": "object",
                "properties": {
                    "query": {
                        "type": "string",
                        "description": "What to search for"
                    }
                },
                "required": ["query"]
            }
        }
    }
]

# When the agent calls a tool, route to the REST API:
def handle_tool_call(name, args):
    import httpx
    base = "https://your-forge-url"
    headers = {"x-connect-token": "YOUR_TOKEN", "Content-Type": "application/json"}

    if name == "store_memory":
        r = httpx.post(f"{base}/api/v1/memory/commit",
                       headers=headers,
                       json={"content": args["content"], "agentId": "my-agent"})
        return r.json()
    elif name == "recall_memory":
        r = httpx.post(f"{base}/api/v1/sdk/recall",
                       headers=headers,
                       json={"query": args["query"], "agentId": "my-agent"})
        return r.json()

Other Frameworks

The same REST API pattern works with any agent framework. For dedicated integrations with richer features, see: