agent-framework-harny 0.1.0b1__py3-none-any.whl

agent_framework_harny/__init__.py

@@ -0,0 +1,21 @@
+"""Harny memory for the Microsoft Agent Framework."""
+
+from ._client import (
+    AsyncHarnyMemoryClient,
+    HarnyAPIError,
+    HarnyMemoryClient,
+    resolve_base_url,
+)
+from ._provider import STATE_USER_KEY, HarnyContextProvider
+
+__version__ = "0.1.0b1"
+
+__all__ = [
+    "AsyncHarnyMemoryClient",
+    "HarnyAPIError",
+    "HarnyContextProvider",
+    "HarnyMemoryClient",
+    "STATE_USER_KEY",
+    "__version__",
+    "resolve_base_url",
+]

agent_framework_harny/_client.py

@@ -0,0 +1,341 @@
+"""A small client for Harny's memory API (`/v1/memory`): send a fact, wait for it, recall, forget.
+
+It is deliberately thin: the field names are the API's own (camelCase on the wire, snake_case in Python arguments) and the
+answers come back as plain dicts. Nothing here logs content or keys.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import os
+import time
+from collections.abc import Iterable, Mapping
+from typing import Any
+from urllib.parse import quote, urlsplit
+
+import httpx
+
+DEFAULT_BASE_URL = "https://api.harny.ai"
+RECALL_LEVELS = ("low", "medium", "high", "extra-high")
+_LOOPBACK_HOSTS = {"localhost", "127.0.0.1", "::1"}
+_USER_KEY_PREFIX = "fsk_"
+
+
+class HarnyAPIError(Exception):
+    """The memory API refused a call, or could not be reached.
+
+    ``code`` is the API's own error code (``unauthorized``, ``recall-query-required``, ``project-cap-reached`` ...), or
+    ``network`` / ``unreadable-response`` when no API answer exists. A per-minute rate limit has no body: ``code`` is
+    ``rate-limited`` and ``retry_after`` carries the seconds to wait.
+    """
+
+    def __init__(
+        self,
+        code: str,
+        *,
+        status: int | None = None,
+        debug_message: str = "",
+        retry_after: float | None = None,
+        data: Any = None,
+    ) -> None:
+        super().__init__(f"{code}" + (f" ({status})" if status else "") + (f": {debug_message}" if debug_message else ""))
+        self.code = code
+        self.status = status
+        self.debug_message = debug_message
+        self.retry_after = retry_after
+        self.data = data
+
+
+def resolve_base_url(base_url: str | None = None) -> str:
+    """The API root with ``/v1`` on the end. HTTPS only; plain HTTP is accepted for a loopback host (local testing)."""
+    raw = (base_url or os.environ.get("HARNY_BASE_URL") or DEFAULT_BASE_URL).strip().rstrip("/")
+    parts = urlsplit(raw)
+    host = (parts.hostname or "").lower()
+    if parts.scheme == "https" and host:
+        pass
+    elif parts.scheme == "http" and (host in _LOOPBACK_HOSTS or host.endswith(".localhost")):
+        pass
+    else:
+        raise ValueError("base_url must be an https:// address (http:// is accepted only for localhost)")
+    if parts.query or parts.fragment or parts.username or parts.password:
+        raise ValueError("base_url must not carry a query, a fragment or credentials")
+    path = parts.path.rstrip("/")
+    if not path.endswith("/v1"):
+        path += "/v1"
+    return f"{parts.scheme}://{parts.netloc}{path}"
+
+
+def resolve_api_key(api_key: Any = None) -> str:
+    """A key given directly (str or a ``SecretString``), else ``HARNY_API_KEY``. Never echoes the key."""
+    if api_key is not None and hasattr(api_key, "get_secret_value"):
+        api_key = api_key.get_secret_value()
+    key = (api_key or os.environ.get("HARNY_API_KEY") or "").strip()
+    if not key:
+        raise ValueError(
+            "No Harny API key. Pass api_key=... or set HARNY_API_KEY. "
+            'To make an app and save its key to .env without printing it, run: harny init --name "<business>"'
+        )
+    return key
+
+
+def _headers(api_key: str, user: str | None) -> dict[str, str]:
+    headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json", "Accept": "application/json"}
+    # An app key (fsa_) acts for a person named in X-Harny-User; a person's own key (fsk_) already is that person.
+    if user and not api_key.startswith(_USER_KEY_PREFIX):
+        headers["X-Harny-User"] = user
+    return headers
+
+
+def _error_from(response: httpx.Response) -> HarnyAPIError:
+    retry_after: float | None = None
+    raw = response.headers.get("Retry-After")
+    if raw:
+        try:
+            retry_after = float(raw)
+        except ValueError:
+            retry_after = None
+    body: Any = None
+    if response.content:
+        try:
+            body = response.json()
+        except ValueError:
+            body = None
+    if isinstance(body, Mapping) and isinstance(body.get("code"), str):
+        return HarnyAPIError(
+            body["code"],
+            status=response.status_code,
+            debug_message=str(body.get("debugMessage") or ""),
+            retry_after=retry_after,
+            data=body.get("data"),
+        )
+    if response.status_code == 429:
+        return HarnyAPIError("rate-limited", status=429, retry_after=retry_after)
+    return HarnyAPIError("unreadable-response", status=response.status_code)
+
+
+def _json(response: httpx.Response) -> Any:
+    if response.status_code >= 400:
+        raise _error_from(response)
+    if not response.content:
+        return None
+    try:
+        return response.json()
+    except ValueError as exc:
+        raise HarnyAPIError("unreadable-response", status=response.status_code) from exc
+
+
+def _ingest_body(content: str, id: str | None, label: str | None, timestamp: str | None, metadata: Mapping[str, Any] | None) -> dict[str, Any]:
+    if not isinstance(content, str) or not content.strip():
+        raise ValueError("content must be a non-blank string")
+    if id is not None and (not id or "/" in id):
+        raise ValueError("id must be a non-empty string without '/'")
+    body: dict[str, Any] = {"content": content}
+    if id is not None:
+        body["id"] = id
+    if label is not None:
+        body["label"] = label
+    if timestamp is not None:
+        body["timestamp"] = timestamp
+    if metadata is not None:
+        body["metadata"] = dict(metadata)
+    return body
+
+
+def _recall_body(
+    query: str, identity: str, level: str | None, max_tokens: int | None, limit: int | None, sort: str | None
+) -> dict[str, Any]:
+    if not isinstance(query, str) or not query.strip():
+        raise ValueError("query must be a non-blank string")
+    if not identity:
+        raise ValueError("identity is required: the person whose memory is read")
+    if level is not None and max_tokens is not None:
+        raise ValueError("pass level or max_tokens, not both")
+    if level is not None and level not in RECALL_LEVELS:
+        raise ValueError(f"level must be one of {', '.join(RECALL_LEVELS)}")
+    body: dict[str, Any] = {"query": query, "identity": identity}
+    if level is not None:
+        body["level"] = level
+    if max_tokens is not None:
+        body["maxTokens"] = max_tokens
+    if limit is not None:
+        body["limit"] = limit
+    if sort is not None:
+        body["sort"] = sort
+    return body
+
+
+def _delete_body(ids: Iterable[str]) -> dict[str, Any]:
+    unique = list(dict.fromkeys(ids))
+    if not unique:
+        raise ValueError("ids must hold at least one source id")
+    return {"ids": unique}
+
+
+def _status_path(ingestion_id: str) -> str:
+    if not ingestion_id:
+        raise ValueError("ingestion_id must be non-empty")
+    return f"/memory/ingest/{quote(ingestion_id, safe='')}"
+
+
+class _Base:
+    def __init__(self, api_key: Any, base_url: str | None, user: str | None, timeout_s: float) -> None:
+        self._api_key = resolve_api_key(api_key)
+        self.base_url = resolve_base_url(base_url)
+        self.user = user
+        self._timeout_s = timeout_s
+
+    def _url(self, path: str) -> str:
+        return f"{self.base_url}{path}"
+
+    def _h(self, user: str | None = None) -> dict[str, str]:
+        return _headers(self._api_key, user or self.user)
+
+
+class AsyncHarnyMemoryClient(_Base):
+    """Async client. Pass ``http_client`` (an ``httpx.AsyncClient``) to bring your own transport; it is not closed."""
+
+    def __init__(
+        self,
+        api_key: Any = None,
+        *,
+        base_url: str | None = None,
+        user: str | None = None,
+        http_client: httpx.AsyncClient | None = None,
+        timeout_s: float = 20.0,
+    ) -> None:
+        super().__init__(api_key, base_url, user, timeout_s)
+        self._owns_client = http_client is None
+        self._http = http_client or httpx.AsyncClient(timeout=timeout_s)
+
+    async def _send(self, method: str, path: str, body: Any = None, user: str | None = None) -> Any:
+        try:
+            response = await self._http.request(method, self._url(path), headers=self._h(user), json=body, timeout=self._timeout_s)
+        except httpx.HTTPError as exc:
+            raise HarnyAPIError("network", debug_message=type(exc).__name__) from exc
+        return _json(response)
+
+    async def ingest(
+        self,
+        content: str,
+        *,
+        id: str | None = None,  # noqa: A002 - the API's own field name
+        label: str | None = None,
+        timestamp: str | None = None,
+        metadata: Mapping[str, Any] | None = None,
+        user: str | None = None,
+    ) -> dict[str, Any]:
+        return await self._send("POST", "/memory/ingest", _ingest_body(content, id, label, timestamp, metadata), user)
+
+    async def get_ingestion(self, ingestion_id: str, *, user: str | None = None) -> dict[str, Any]:
+        return await self._send("GET", _status_path(ingestion_id), None, user)
+
+    async def wait_for_ingestion(
+        self, ingestion_id: str, *, timeout_s: float = 120.0, poll_s: float = 2.0, user: str | None = None
+    ) -> dict[str, Any]:
+        deadline = time.monotonic() + timeout_s
+        while True:
+            state = await self.get_ingestion(ingestion_id, user=user)
+            if state.get("status") != "processing" or time.monotonic() >= deadline:
+                return state
+            await asyncio.sleep(poll_s)
+
+    async def recall(
+        self,
+        query: str,
+        *,
+        identity: str,
+        level: str | None = None,
+        max_tokens: int | None = None,
+        limit: int | None = None,
+        sort: str | None = None,
+        user: str | None = None,
+    ) -> dict[str, Any]:
+        return await self._send("POST", "/memory/recall", _recall_body(query, identity, level, max_tokens, limit, sort), user)
+
+    async def delete_data_points(self, ids: Iterable[str], *, user: str | None = None) -> dict[str, Any]:
+        return await self._send("POST", "/memory/data-points/delete", _delete_body(ids), user)
+
+    async def aclose(self) -> None:
+        if self._owns_client:
+            await self._http.aclose()
+
+    async def __aenter__(self) -> "AsyncHarnyMemoryClient":
+        return self
+
+    async def __aexit__(self, *exc_info: object) -> None:
+        await self.aclose()
+
+
+class HarnyMemoryClient(_Base):
+    """Sync client with the same methods. Pass ``http_client`` (an ``httpx.Client``) to bring your own transport."""
+
+    def __init__(
+        self,
+        api_key: Any = None,
+        *,
+        base_url: str | None = None,
+        user: str | None = None,
+        http_client: httpx.Client | None = None,
+        timeout_s: float = 20.0,
+    ) -> None:
+        super().__init__(api_key, base_url, user, timeout_s)
+        self._owns_client = http_client is None
+        self._http = http_client or httpx.Client(timeout=timeout_s)
+
+    def _send(self, method: str, path: str, body: Any = None, user: str | None = None) -> Any:
+        try:
+            response = self._http.request(method, self._url(path), headers=self._h(user), json=body, timeout=self._timeout_s)
+        except httpx.HTTPError as exc:
+            raise HarnyAPIError("network", debug_message=type(exc).__name__) from exc
+        return _json(response)
+
+    def ingest(
+        self,
+        content: str,
+        *,
+        id: str | None = None,  # noqa: A002
+        label: str | None = None,
+        timestamp: str | None = None,
+        metadata: Mapping[str, Any] | None = None,
+        user: str | None = None,
+    ) -> dict[str, Any]:
+        return self._send("POST", "/memory/ingest", _ingest_body(content, id, label, timestamp, metadata), user)
+
+    def get_ingestion(self, ingestion_id: str, *, user: str | None = None) -> dict[str, Any]:
+        return self._send("GET", _status_path(ingestion_id), None, user)
+
+    def wait_for_ingestion(
+        self, ingestion_id: str, *, timeout_s: float = 120.0, poll_s: float = 2.0, user: str | None = None
+    ) -> dict[str, Any]:
+        deadline = time.monotonic() + timeout_s
+        while True:
+            state = self.get_ingestion(ingestion_id, user=user)
+            if state.get("status") != "processing" or time.monotonic() >= deadline:
+                return state
+            time.sleep(poll_s)
+
+    def recall(
+        self,
+        query: str,
+        *,
+        identity: str,
+        level: str | None = None,
+        max_tokens: int | None = None,
+        limit: int | None = None,
+        sort: str | None = None,
+        user: str | None = None,
+    ) -> dict[str, Any]:
+        return self._send("POST", "/memory/recall", _recall_body(query, identity, level, max_tokens, limit, sort), user)
+
+    def delete_data_points(self, ids: Iterable[str], *, user: str | None = None) -> dict[str, Any]:
+        return self._send("POST", "/memory/data-points/delete", _delete_body(ids), user)
+
+    def close(self) -> None:
+        if self._owns_client:
+            self._http.close()
+
+    def __enter__(self) -> "HarnyMemoryClient":
+        return self
+
+    def __exit__(self, *exc_info: object) -> None:
+        self.close()

agent_framework_harny/_provider.py

@@ -0,0 +1,256 @@
+"""``HarnyContextProvider``: Harny memory for a Microsoft Agent Framework agent.
+
+Before each run it recalls what the person's memory holds for what they just said and hands it to the model. After
+each run it saves what the person said. A memory problem never fails the agent run.
+"""
+
+from __future__ import annotations
+
+import hashlib
+import logging
+from typing import TYPE_CHECKING, Any
+
+import httpx
+from agent_framework import ContextProvider, Message, tool
+from agent_framework._sessions import AgentSession, SessionContext
+
+from ._client import AsyncHarnyMemoryClient, HarnyAPIError
+
+if TYPE_CHECKING:  # pragma: no cover
+    from agent_framework._agents import SupportsAgentRun
+
+logger = logging.getLogger("agent_framework_harny")
+
+STATE_USER_KEY = "harny_user_id"
+_MEMORY_LABEL = "Agent conversation"
+
+
+def _role_value(role: Any) -> str:
+    return str(getattr(role, "value", role))
+
+
+class HarnyContextProvider(ContextProvider):
+    """Gives an agent long-term memory of each person it talks to, backed by Harny.
+
+    Args:
+        source_id: Name of this provider on the agent (default ``harny``).
+        api_key: A Harny app key (``fsa_...``) or a person's own key (``fsk_...``), as text or a ``SecretString``.
+            Default: the ``HARNY_API_KEY`` environment variable.
+        base_url: The API address. Default ``https://api.harny.ai`` (``/v1`` is added when missing). HTTPS only;
+            plain HTTP is accepted for localhost.
+        user_id: Your id for the person this agent talks to. With an app key it is sent as ``X-Harny-User``. When it
+            is ``None``, the provider reads ``session.state["harny_user_id"]`` on each run, so one provider can serve
+            many people. With no id at all, memory is skipped for that run (a warning is logged once).
+        context_prompt: The heading put above the recalled facts.
+        level: How much evidence to recall: ``low``, ``medium`` (default), ``high`` or ``extra-high``.
+        max_tokens: An explicit evidence budget instead of ``level``.
+        limit: The most facts to recall (1 to 200).
+        save_user_messages: Save what the person says after each run. Off by default; pass ``True`` to switch it
+            on. Only the person's own messages are saved, never the assistant's or the system's.
+        expose_tools: Also give the agent two tools it can call: ``harny_recall`` and ``harny_remember``.
+        http_client: Your own ``httpx.AsyncClient`` (for a proxy or a test transport). It is not closed by the provider.
+    """
+
+    DEFAULT_CONTEXT_PROMPT = "## Memories\nConsider the following memories about this person when answering:"
+    DEFAULT_SOURCE_ID = "harny"
+
+    def __init__(
+        self,
+        source_id: str = DEFAULT_SOURCE_ID,
+        *,
+        api_key: Any = None,
+        base_url: str | None = None,
+        user_id: str | None = None,
+        context_prompt: str | None = None,
+        level: str | None = "medium",
+        max_tokens: int | None = None,
+        limit: int | None = None,
+        save_user_messages: bool = False,
+        expose_tools: bool = False,
+        http_client: httpx.AsyncClient | None = None,
+    ) -> None:
+        super().__init__(source_id)
+        if max_tokens is not None:
+            level = None
+        self.user_id = user_id
+        self.context_prompt = context_prompt or self.DEFAULT_CONTEXT_PROMPT
+        self.level = level
+        self.max_tokens = max_tokens
+        self.limit = limit
+        self.save_user_messages = save_user_messages
+        self.expose_tools = expose_tools
+        self.client = AsyncHarnyMemoryClient(api_key, base_url=base_url, http_client=http_client)
+        self._warned_no_user = False
+        self._pending_ingestions: list[tuple[str, str]] = []
+
+    # -- lifecycle -----------------------------------------------------------------------------------------------
+
+    async def __aenter__(self) -> "HarnyContextProvider":
+        return self
+
+    async def __aexit__(self, *exc_info: object) -> None:
+        await self.client.aclose()
+
+    # -- hooks ---------------------------------------------------------------------------------------------------
+
+    def _user_for(self, session: AgentSession | None) -> str | None:
+        if self.user_id:
+            return self.user_id
+        state = getattr(session, "state", None)
+        value = state.get(STATE_USER_KEY) if isinstance(state, dict) else None
+        if isinstance(value, str) and value.strip():
+            return value.strip()
+        if not self._warned_no_user:
+            self._warned_no_user = True
+            logger.warning(
+                "HarnyContextProvider has no person to remember: pass user_id=... or set session.state['%s']. "
+                "Memory is skipped.",
+                STATE_USER_KEY,
+            )
+        return None
+
+    async def before_run(
+        self,
+        *,
+        agent: "SupportsAgentRun",
+        session: AgentSession,
+        context: SessionContext,
+        state: dict[str, Any],
+    ) -> None:
+        if self.expose_tools:
+            context.extend_tools(self.source_id, self._tools(session))
+        user = self._user_for(session)
+        if not user:
+            return
+        query = "\n".join(
+            m.text for m in context.input_messages if m and _role_value(m.role) == "user" and m.text and m.text.strip()
+        )
+        if not query.strip():
+            return
+        try:
+            answer = await self.client.recall(
+                query, identity=user, level=self.level, max_tokens=self.max_tokens, limit=self.limit, user=user
+            )
+        except (HarnyAPIError, ValueError) as exc:
+            logger.warning("Harny recall failed (%s); the agent runs without memory.", getattr(exc, "code", "invalid-request"))
+            return
+        text = self._render(answer)
+        if text:
+            context.extend_messages(self.source_id, [Message(role="user", contents=[text])])
+
+    async def after_run(
+        self,
+        *,
+        agent: "SupportsAgentRun",
+        session: AgentSession,
+        context: SessionContext,
+        state: dict[str, Any],
+    ) -> None:
+        if not self.save_user_messages:
+            return
+        user = self._user_for(session)
+        if not user:
+            return
+        session_id = context.session_id or getattr(session, "session_id", None) or ""
+        for message in context.input_messages:
+            if _role_value(message.role) != "user" or not message.text or not message.text.strip():
+                continue
+            try:
+                receipt = await self.client.ingest(
+                    message.text,
+                    id=self._data_point_id(session_id, message.text),
+                    label=_MEMORY_LABEL,
+                    metadata={"source": "agent-framework", "role": "user"},
+                    user=user,
+                )
+            except (HarnyAPIError, ValueError) as exc:
+                logger.warning("Harny save failed (%s); the answer is unaffected.", getattr(exc, "code", "invalid-request"))
+                return
+            ingestion_id = receipt.get("ingestionId") if isinstance(receipt, dict) else None
+            if isinstance(ingestion_id, str):
+                self._pending_ingestions.append((user, ingestion_id))
+
+    # -- helpers -------------------------------------------------------------------------------------------------
+
+    @staticmethod
+    def _data_point_id(session_id: str, text: str) -> str:
+        """The same person saying the same words in the same session is one fact, however often it is sent."""
+        return hashlib.sha256(f"{session_id}\n{text}".encode()).hexdigest()
+
+    def _render(self, answer: Any) -> str:
+        if not isinstance(answer, dict):
+            return ""
+        lines: list[str] = []
+        for result in sorted(answer.get("results") or [], key=lambda r: r.get("rank", 0)):
+            content = str(result.get("content") or "").strip()
+            if not content:
+                continue
+            when = str(result.get("occurredAt") or "")[:10]
+            lines.append(f"- {content}" + (f" ({when})" if when else ""))
+        pending = answer.get("pending")
+        if isinstance(pending, dict) and pending.get("dataPoints"):
+            lines.append(
+                f"(Note: {pending['dataPoints']} recent fact(s) are still being processed, so the newest "
+                "information may be missing.)"
+            )
+        if not any(line.startswith("- ") for line in lines):
+            return ""
+        return f"{self.context_prompt}\n" + "\n".join(lines)
+
+    async def wait_until_ready(self, *, timeout_s: float = 120.0, poll_s: float = 2.0) -> bool:
+        """Wait until everything this provider saved has been processed. True when nothing failed or timed out."""
+        ok = True
+        waiting, self._pending_ingestions = self._pending_ingestions, []
+        for user, ingestion_id in waiting:
+            try:
+                state = await self.client.wait_for_ingestion(ingestion_id, timeout_s=timeout_s, poll_s=poll_s, user=user)
+            except HarnyAPIError as exc:
+                logger.warning("Harny status check failed (%s).", exc.code)
+                ok = False
+                continue
+            if state.get("status") != "completed":
+                ok = False
+        return ok
+
+    def _tools(self, session: AgentSession) -> list[Any]:
+        """Two tools the agent may call. They act for the same person as the provider and never raise to the model."""
+        user = self._user_for(session)
+
+        @tool(
+            name="harny_recall",
+            description="Look up what is remembered about this person. Pass a short question, e.g. 'What is the Acme pilot budget?'.",
+        )
+        async def harny_recall(query: str) -> str:
+            if not user:
+                return "Memory is not available for this conversation."
+            try:
+                answer = await self.client.recall(
+                    query, identity=user, level=self.level, max_tokens=self.max_tokens, limit=self.limit, user=user
+                )
+            except (HarnyAPIError, ValueError) as exc:
+                return f"Memory is unavailable right now ({getattr(exc, 'code', 'invalid-request')})."
+            return self._render(answer) or "Nothing is remembered about that yet."
+
+        @tool(
+            name="harny_remember",
+            description="Remember a fact about this person for later conversations. Pass one clear sentence.",
+        )
+        async def harny_remember(fact: str) -> str:
+            if not user:
+                return "Memory is not available for this conversation."
+            try:
+                receipt = await self.client.ingest(
+                    fact,
+                    id=self._data_point_id("tool", fact),
+                    label="Saved by the agent",
+                    metadata={"source": "agent-framework", "role": "agent"},
+                    user=user,
+                )
+            except (HarnyAPIError, ValueError) as exc:
+                return f"Could not save that right now ({getattr(exc, 'code', 'invalid-request')})."
+            ingestion_id = receipt.get("ingestionId") if isinstance(receipt, dict) else None
+            if isinstance(ingestion_id, str):
+                self._pending_ingestions.append((user, ingestion_id))
+            return "Saved."
+
+        return [harny_recall, harny_remember]

agent_framework_harny-0.1.0b1.dist-info/METADATA

@@ -0,0 +1,149 @@
+Metadata-Version: 2.4
+Name: agent-framework-harny
+Version: 0.1.0b1
+Summary: Long-term memory for Microsoft Agent Framework agents: a context provider backed by Harny.
+Author-email: Harny <platform@harny.ai>
+License: Apache-2.0
+Project-URL: Homepage, https://harny.ai
+Project-URL: Documentation, https://harny.ai/docs
+Keywords: harny,agent-framework,memory,context-provider,ai-agents
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Typing :: Typed
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: agent-framework-core<2,>=1.13.0
+Requires-Dist: httpx<1.0.0,>=0.27.2
+Provides-Extra: dev
+Requires-Dist: pytest>=8.3.0; extra == "dev"
+Requires-Dist: pytest-asyncio>=0.24.0; extra == "dev"
+Requires-Dist: pytest-socket>=0.7.0; extra == "dev"
+Dynamic: license-file
+
+# agent-framework-harny
+
+Long-term memory for Microsoft Agent Framework agents. `HarnyContextProvider` recalls what each person's memory holds before every run (and, when you switch saving on, saves what they say after it), so your agent remembers across sessions, restarts and devices. Pre-release (`0.1.0b1`), in the same maturity band as the framework's own integrations.
+
+## Install
+
+```bash
+pip install agent-framework-harny --pre
+```
+
+Python 3.10+. Needs `agent-framework-core>=1.13,<2` (tested against 1.23.0).
+
+## Start without a key
+
+You do not need an account first. From the project folder:
+
+```bash
+pip install harny
+harny init --name "<business>" --about "<what the app does>" --customers "<who its customers are>" --preset cost_first
+```
+
+That creates the app and saves `HARNY_API_KEY` to `.env` without printing it. Never print, log or commit the key. (Raw API route: `POST https://api.harny.ai/v1/auth/provision` with `{"app_name": "<business>"}`; the key is shown once.)
+
+## One-minute example
+
+```python
+import asyncio
+
+from agent_framework import Agent
+from agent_framework_harny import HarnyContextProvider
+
+client = ...  # any Agent Framework chat client you already use
+
+
+async def main() -> None:
+    async with HarnyContextProvider(user_id="nadia", save_user_messages=True) as memory:  # reads HARNY_API_KEY
+        agent = Agent(client, "You help an account manager.", context_providers=[memory])
+        session = agent.create_session()
+        await agent.run("Nadia raised the Acme pilot budget from 32k to 40k.", session=session)
+        await memory.wait_until_ready()  # facts are processed in the background; demos and tests wait, apps need not
+        new_session = agent.create_session()  # a different session: only memory carries the fact over
+        print((await agent.run("What is the Acme pilot budget?", session=new_session)).text)
+
+
+asyncio.run(main())
+```
+
+Saving conversation messages is off by default; the examples switch it on with `save_user_messages=True`. `user_id` is your own id for the person. One provider can serve many people: leave `user_id` out and set `session.state["harny_user_id"] = "<id>"` on each session.
+
+## The canonical example, as an agent
+
+Same facts as every Harny surface (Nadia, the Acme pilot, the budget that moved from 32k to 40k):
+
+```python
+async with HarnyContextProvider(user_id="nadia", save_user_messages=True) as memory:
+    agent = Agent(client, "You help an account manager.", context_providers=[memory])
+    session = agent.create_session()
+    await agent.run("Nadia confirmed the Acme pilot ships March 14. Budget is 32k.", session=session)
+    await agent.run("Nadia raised the Acme pilot budget from 32k to 40k.", session=session)
+    await memory.wait_until_ready()
+    reply = await agent.run("What is the Acme pilot budget?", session=agent.create_session())
+    print(reply.text)  # the model sees: "Acme pilot budget: currently 40k."
+```
+
+## Options
+
+| Argument | Default | Meaning |
+|---|---|---|
+| `api_key` | `HARNY_API_KEY` | App key (`fsa_`) or a person's own key (`fsk_`); text or `SecretString` |
+| `user_id` | `None` | Your id for the person. Needed with an app key (sent as `X-Harny-User`) |
+| `base_url` | `https://api.harny.ai` | `/v1` is added when missing. HTTPS only; `http://localhost` allowed for local runs |
+| `level` / `max_tokens` | `medium` | How much evidence to recall (`low`, `medium`, `high`, `extra-high`, or a token budget) |
+| `limit` | API default | Most facts recalled (1 to 200) |
+| `save_user_messages` | `False` | Off by default. `True` saves what the person says after each run |
+| `expose_tools` | `False` | Also give the agent `harny_recall` and `harny_remember` tools |
+| `context_prompt` | "## Memories ..." | Heading above the recalled facts |
+| `http_client` | new client | Your own `httpx.AsyncClient` (proxy, test transport); not closed by the provider |
+
+## Behaviour an agent should know
+
+- **Saving conversation messages is off by default** (`save_user_messages=True` switches it on). When on, only the person's own messages are saved, never the assistant's replies or system text. `expose_tools=True` lets the agent save a fact on request.
+- **A saved fact has a stable id** (hash of the session id and the words), so sending the same words in the same session twice is one fact and costs nothing the second time.
+- **A memory problem never fails the run.** If recall or save is refused, the agent answers without memory and one warning with the error code is logged (never content, never keys).
+- **`pending`**: facts saved a moment ago may still be processing. The model is told, in one line, that the newest information may be missing.
+- **Errors** are `HarnyAPIError` with `code`, `status`, `debug_message`, `retry_after` (a per-minute rate limit has no body; `code` is `rate-limited`).
+- Forget a fact: `await memory.client.delete_data_points([source_id])`.
+
+## The memory API underneath
+
+The provider is a thin layer over Harny's public memory API (`/v1/memory`: ingest, status, recall, delete). The same calls from Python with the Harny SDK:
+
+```python
+from harny import SyncHarnyClient
+
+with SyncHarnyClient() as harny:  # reads HARNY_API_KEY
+    harny.ingest_data_point(
+        "Nadia confirmed the Acme pilot ships March 14. Budget is 32k.",
+        id="call-8820", label="Account review call", timestamp="2026-02-03T09:00:00Z", user="nadia",
+    )
+    receipt = harny.ingest_data_point(
+        "Nadia raised the Acme pilot budget from 32k to 40k.",
+        id="call-8821", label="Budget change", timestamp="2026-03-01T09:00:00Z", user="nadia",
+    )
+    harny.wait_for_ingestion(receipt.ingestionId, user="nadia")
+
+    recall = harny.recall("What is the Acme pilot budget?", level="medium", user="nadia")
+    for result in recall.results:
+        print(result.rank, result.content)
+```
+
+## Tests and pinned versions
+
+```bash
+scripts/ci.sh            # fresh venv, installs the pins in ci-pins.txt, runs the tests (no network, no paid call)
+scripts/ci.sh --latest   # the same tests against the newest framework release (exit 3 = a framework update breaks us)
+```
+
+Pinned: `agent-framework-core==1.23.0`, `httpx==0.28.1`.
+
+Apache-2.0. https://harny.ai

agent_framework_harny-0.1.0b1.dist-info/RECORD

@@ -0,0 +1,9 @@
+agent_framework_harny/__init__.py,sha256=nTO82pZOOVIzPAtpXqvw5uMemVcx_Ggf4ICv-XHqcmI,441
+agent_framework_harny/_client.py,sha256=ThLp0FXsa3OMM-N5tQdXP7BMvWw_EZEEQN7KP9wwHp0,13019
+agent_framework_harny/_provider.py,sha256=41xNTk5bROP0Q3dBTRGFxB_riUJYw_rq03wZxNjFg7s,11411
+agent_framework_harny/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+agent_framework_harny-0.1.0b1.dist-info/licenses/LICENSE,sha256=z8d0m5b2O9McPEK1xHG_dWgUBT6EfBDz6wA0F7xSPTA,11358
+agent_framework_harny-0.1.0b1.dist-info/METADATA,sha256=znjj-mfR8U2PE-EABUwtKTNtiMAelhicfsWql2MUw2Q,7380
+agent_framework_harny-0.1.0b1.dist-info/WHEEL,sha256=mOdXR9C3wMVov7cSY8PY-XCzo3SmmWCHvGQr5csJoao,91
+agent_framework_harny-0.1.0b1.dist-info/top_level.txt,sha256=wZ_qOktchFz-C26NC3rzn-ba4YZaCG-FtPCzYSd3uhs,22
+agent_framework_harny-0.1.0b1.dist-info/RECORD,,

agent_framework_harny-0.1.0b1.dist-info/WHEEL

@@ -0,0 +1,5 @@
+Wheel-Version: 1.0
+Generator: setuptools (84.2.1)
+Root-Is-Purelib: true
+Tag: py3-none-any
+

agent_framework_harny-0.1.0b1.dist-info/licenses/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

agent_framework_harny-0.1.0b1.dist-info/top_level.txt

@@ -0,0 +1 @@
+agent_framework_harny