Memory for DSPy agents

• Which value is current. In Harny, a changed value with a newer timestamp becomes the current value, and the earlier value stays answerable by date.
• When a fact happened. Send a timestamp (ISO 8601 with zone) with each fact, and recall can answer as of a date with queryTimestamp.
• One person across sources. Harny keeps each person's memory separate, and you name the person with the X-Harny-User header.

Each tool reads or writes one signed-in user's memory through the Memory API.

Your code binds the signed-in user, so the model never chooses whose memory it reads. The app key stays on the server, and each request names the user from your own session.

# dspy.configure(lm=...) has already set your language model
import os
import requests

BASE = "https://api.harny.ai/v1/memory"

def make_memory_tools(user_id: str):
    headers = {
        "Authorization": f"Bearer {os.environ['HARNY_APP_KEY']}",
        "X-Harny-User": user_id,  # set by your application, never by the model
    }

    def recall_memory(question: str) -> str:
        """Look up what is known about this user that is relevant to a question."""
        r = requests.post(f"{BASE}/recall", headers=headers,
                          json={"query": question, "identity": user_id})
        return r.text

    def store_memory(text: str, happened_at: str) -> str:
        """Save a fact with the ISO time it happened."""
        r = requests.post(f"{BASE}/ingest", headers=headers,
                          json={"content": text, "timestamp": happened_at})
        return r.text

    return [recall_memory, store_memory]

import dspy

react = dspy.ReAct("question -> answer", tools=make_memory_tools(user_id))
print(react(question="What is the Acme pilot budget?").answer)

Recall returns results with sources. Each source has a sourceId, an occurredAt and a label.

• POST https://api.harny.ai/v1/memory/ingest takes id, content, label and timestamp.
• POST https://api.harny.ai/v1/memory/recall takes query and identity, and returns results with sources.

Example: your agent asks “what is the Acme pilot budget?” and gets this back. Your code hands the model the content and the source label, so it can say who said it and when.

{
  "asOf": "2026-07-29T00:00:00Z",
  "results": [
    {
      "rank": 1,
      "occurredAt": "2026-07-28T16:00:00Z",
      "content": "Acme pilot budget: currently 40k.",
      "sources": [
        { "sourceId": "call-8821", "label": "Email from Nadia" }
      ]
    }
  ]
}

• Use DSPy's own state for the program's job: its signatures, modules, and the tool calls within a single run.
• Add Harny when facts about a person need to carry dates and stay separate for each user.