loopgrid-haystack 0.1.0__py3-none-any.whl

loopgrid_haystack/__init__.py

@@ -0,0 +1,15 @@
+from .integration import (
+    HAYSTACK_TARGET_VERSION,
+    LOOPGRID_HAYSTACK_VERSION,
+    LoopGridHaystack,
+    LoopGridSpan,
+    LoopGridTracer,
+)
+
+__all__ = [
+    "HAYSTACK_TARGET_VERSION",
+    "LOOPGRID_HAYSTACK_VERSION",
+    "LoopGridHaystack",
+    "LoopGridSpan",
+    "LoopGridTracer",
+]

loopgrid_haystack/integration.py

@@ -0,0 +1,778 @@
+from __future__ import annotations
+
+import contextlib
+import contextvars
+import hashlib
+import json
+import queue
+import threading
+from collections.abc import Iterator, Mapping, Sequence
+from dataclasses import dataclass, field
+from typing import Any
+
+from haystack import tracing
+from haystack.tracing import Span, Tracer
+from loopgrid import LoopGrid
+
+LOOPGRID_HAYSTACK_VERSION = "0.1.0"
+HAYSTACK_TARGET_VERSION = "3.3.0"
+_FRAMEWORK = "haystack"
+
+
+def _require_nonempty(name: str, value: Any) -> None:
+    if value is None or value == "" or value == {} or value == []:
+        raise ValueError(f"{name} is required")
+
+
+def _canonical_bytes(value: Any) -> bytes:
+    try:
+        return json.dumps(
+            value,
+            sort_keys=True,
+            separators=(",", ":"),
+            ensure_ascii=False,
+            default=str,
+        ).encode("utf-8")
+    except Exception:
+        return repr(value).encode("utf-8", errors="replace")
+
+
+def _commitment(value: Any) -> dict[str, Any]:
+    raw = _canonical_bytes(value)
+    return {
+        "algorithm": "sha256",
+        "sha256": hashlib.sha256(raw).hexdigest(),
+        "bytes": len(raw),
+    }
+
+
+def _idempotency(*parts: Any) -> str:
+    raw = "|".join("" if p is None else str(p) for p in parts).encode("utf-8")
+    return "lg-haystack-" + hashlib.sha256(raw).hexdigest()
+
+
+def _call_identity(tool_call: Any) -> str:
+    call_id = getattr(tool_call, "id", None)
+    name = getattr(tool_call, "tool_name", None) or "tool"
+    arguments = getattr(tool_call, "arguments", None)
+    arguments_hash = _commitment(arguments)["sha256"]
+    if call_id:
+        # Tool-call IDs are provider/framework supplied and are not guaranteed to be unique
+        # across an entire Agent decision. Include the surviving tool name/arguments so a
+        # reused ID with modified parameters cannot collapse correlation within one step.
+        return f"id-{call_id}-{name}-{arguments_hash}"
+    return f"local-{name}-{arguments_hash}"
+
+
+def _tool_step_identity(state: Any) -> str:
+    """Return a stable identity for the current Agent tool step.
+
+    Haystack 3.x exposes ``step_count`` in Agent state. For direct hook/unit usage where a
+    synthetic ``State`` may omit that field, the current conversation length is a deterministic
+    fallback: retrying the same hook against the same pending message keeps the same identity,
+    while a later tool step has additional messages and therefore a different identity.
+    """
+
+    step_count = state.data.get("step_count")
+    if step_count is not None:
+        return f"step-{step_count}"
+    messages = state.data.get("messages") or []
+    return f"messages-{len(messages)}"
+
+
+def _tool_calls_from_state(state: Any) -> list[Any]:
+    messages = state.data.get("messages") or []
+    if not messages:
+        return []
+    calls = getattr(messages[-1], "tool_calls", None)
+    return list(calls or [])
+
+
+def _tool_results_from_state(state: Any) -> list[Any]:
+    results: list[Any] = []
+    for message in state.data.get("messages") or []:
+        result = getattr(message, "tool_call_result", None)
+        if result is not None:
+            results.append(result)
+    return results
+
+
+def _safe_tags(tags: Mapping[str, Any] | None) -> dict[str, Any]:
+    return dict(tags or {})
+
+
+def _model_name_from_tags(tags: Mapping[str, Any], fallback: str | None) -> str | None:
+    preferred = (
+        "haystack.chat_generator.model",
+        "haystack.component.model",
+        "gen_ai.request.model",
+        "model",
+    )
+    for key in preferred:
+        value = tags.get(key)
+        if isinstance(value, (str, int, float)) and value != "":
+            return str(value)
+    for key, value in tags.items():
+        if "model" in str(key).lower() and isinstance(value, (str, int, float)) and value != "":
+            return str(value)
+    return fallback
+
+
+@dataclass(frozen=True)
+class _QueuedEvent:
+    decision_id: str
+    event_type: str
+    payload: dict[str, Any]
+    actor_type: str
+    actor_id: str
+    idempotency_key: str
+
+
+@dataclass(frozen=True)
+class _PendingToolCall:
+    call: Any
+    call_identity: str
+    occurrence_identity: str
+    step_identity: str
+    ordinal: int
+
+
+@dataclass(frozen=True)
+class _PendingToolBatch:
+    decision_id: str
+    calls: tuple[_PendingToolCall, ...]
+    result_offset: int
+
+
+class _EventTransport:
+    """Single FIFO worker so hook/tracer callbacks do not block Haystack execution."""
+
+    def __init__(self, owner: "LoopGridHaystack") -> None:
+        self._owner = owner
+        self._queue: queue.Queue[_QueuedEvent | object] = queue.Queue()
+        self._stop = object()
+        self._worker: threading.Thread | None = None
+        self._worker_lock = threading.Lock()
+
+    def _ensure_worker(self) -> None:
+        with self._worker_lock:
+            if self._worker is not None and self._worker.is_alive():
+                return
+            self._worker = threading.Thread(
+                target=self._worker_main,
+                name="loopgrid-haystack",
+                daemon=True,
+            )
+            self._worker.start()
+
+    def _worker_main(self) -> None:
+        while True:
+            item = self._queue.get()
+            try:
+                if item is self._stop:
+                    return
+                assert isinstance(item, _QueuedEvent)
+                try:
+                    self._owner.client.add_event(
+                        item.decision_id,
+                        item.event_type,
+                        item.payload,
+                        actor_type=item.actor_type,
+                        actor_id=item.actor_id,
+                        idempotency_key=item.idempotency_key,
+                    )
+                except Exception as exc:  # observability/evidence must not break the Agent
+                    self._owner._record_error(item.event_type, item.decision_id, exc)
+            finally:
+                self._queue.task_done()
+
+    def enqueue(
+        self,
+        decision_id: str,
+        event_type: str,
+        payload: dict[str, Any],
+        *,
+        actor_type: str,
+        actor_id: str,
+        identity: Sequence[Any],
+    ) -> None:
+        self._ensure_worker()
+        self._queue.put(
+            _QueuedEvent(
+                decision_id=decision_id,
+                event_type=event_type,
+                payload=payload,
+                actor_type=actor_type,
+                actor_id=actor_id,
+                idempotency_key=_idempotency(decision_id, event_type, *identity),
+            )
+        )
+
+    def flush(self) -> None:
+        self._ensure_worker()
+        self._queue.join()
+
+    def shutdown(self) -> None:
+        self.flush()
+        worker = self._worker
+        if worker is not None and worker.is_alive():
+            self._queue.put(self._stop)
+            self._queue.join()
+            worker.join(timeout=5)
+
+
+@dataclass
+class LoopGridSpan(Span):
+    """Haystack Span implementation used by :class:`LoopGridTracer`."""
+
+    owner: "LoopGridHaystack"
+    operation_name: str
+    decision_id: str | None
+    parent_operation: str | None = None
+    tags: dict[str, Any] = field(default_factory=dict)
+    delegate_span: Span | None = None
+    sequence: int | None = None
+
+    def set_tag(self, key: str, value: Any) -> None:
+        self.tags[key] = value
+        if self.delegate_span is not None:
+            try:
+                self.delegate_span.set_tag(key, value)
+            except Exception as exc:
+                self.owner._record_error("delegate_set_tag", self.decision_id, exc)
+
+
+class LoopGridTracer(Tracer):
+    """Native Haystack tracer that records LLM-completion evidence.
+
+    Tool request/execution evidence is intentionally *not* inferred from tracing spans. It is
+    recorded by Agent hooks around Haystack's actual tool execution boundary so approval-gated
+    or rejected calls are not misreported as executed actions.
+
+    A delegate tracer may be supplied so LoopGrid can coexist with an existing Haystack tracer.
+    """
+
+    def __init__(self, owner: "LoopGridHaystack", delegate: Tracer | None = None) -> None:
+        self._owner = owner
+        self._delegate = delegate
+        self._current: contextvars.ContextVar[LoopGridSpan | None] = contextvars.ContextVar(
+            "loopgrid_haystack_current_span", default=None
+        )
+
+    @contextlib.contextmanager
+    def trace(
+        self,
+        operation_name: str,
+        tags: dict[str, Any] | None = None,
+        parent_span: Span | None = None,
+    ) -> Iterator[Span]:
+        delegate_parent = parent_span.delegate_span if isinstance(parent_span, LoopGridSpan) else None
+        delegate_span: Span | None = None
+        error: BaseException | None = None
+
+        with contextlib.ExitStack() as stack:
+            if self._delegate is not None:
+                try:
+                    delegate_span = stack.enter_context(
+                        self._delegate.trace(operation_name, tags=tags, parent_span=delegate_parent)
+                    )
+                except Exception as exc:
+                    self._owner._record_error("delegate_trace_start", self._owner.current_decision_id, exc)
+                    delegate_span = None
+
+            parent_operation = parent_span.operation_name if isinstance(parent_span, LoopGridSpan) else None
+            decision_id = self._owner.current_decision_id
+            span = LoopGridSpan(
+                owner=self._owner,
+                operation_name=operation_name,
+                decision_id=decision_id,
+                parent_operation=parent_operation,
+                tags=_safe_tags(tags),
+                delegate_span=delegate_span,
+                sequence=self._owner._next_span_sequence(decision_id) if decision_id else None,
+            )
+            token = self._current.set(span)
+            try:
+                yield span
+            except BaseException as exc:  # preserve application exception behavior
+                error = exc
+                raise
+            finally:
+                self._current.reset(token)
+                try:
+                    self._owner._finish_span(span, error)
+                except Exception as exc:
+                    self._owner._record_error("trace_finish", span.decision_id, exc)
+
+    def current_span(self) -> Span | None:
+        return self._current.get()
+
+
+class _BeforeRunHook:
+    allowed_hook_points = frozenset({"before_run"})
+
+    def __init__(self, owner: "LoopGridHaystack", decision_id: str) -> None:
+        self._owner = owner
+        self._decision_id = decision_id
+
+    def run(self, state: Any) -> None:
+        self._owner._enter_run(state, self._decision_id)
+
+    async def run_async(self, state: Any) -> None:
+        self.run(state)
+
+
+class _BeforeToolHook:
+    allowed_hook_points = frozenset({"before_tool"})
+
+    def __init__(self, owner: "LoopGridHaystack", decision_id: str) -> None:
+        self._owner = owner
+        self._decision_id = decision_id
+
+    def run(self, state: Any) -> None:
+        # This hook MUST be registered after ConfirmationHook (or any other hook that can
+        # remove/modify tool calls). We record only calls that survive those controls.
+        calls = _tool_calls_from_state(state)
+        step_identity = _tool_step_identity(state)
+        pending_calls = [
+            _PendingToolCall(
+                call=call,
+                call_identity=_call_identity(call),
+                occurrence_identity=f"{step_identity}:{ordinal}:{_call_identity(call)}",
+                step_identity=step_identity,
+                ordinal=ordinal,
+            )
+            for ordinal, call in enumerate(calls)
+        ]
+        self._owner._set_pending(state, self._decision_id, pending_calls)
+        for pending in pending_calls:
+            call = pending.call
+            name = str(getattr(call, "tool_name", None) or "tool")
+            arguments = getattr(call, "arguments", None)
+            payload: dict[str, Any] = {
+                "framework": _FRAMEWORK,
+                "haystack_target": HAYSTACK_TARGET_VERSION,
+                "tool": name,
+                "call_id": getattr(call, "id", None),
+                "agent_step": pending.step_identity,
+                "tool_ordinal": pending.ordinal,
+                "input_commitment": _commitment(arguments),
+                "source": "agent.before_tool",
+            }
+            if self._owner.capture_content:
+                payload["input"] = arguments
+            self._owner._transport.enqueue(
+                self._decision_id,
+                "tool_requested",
+                payload,
+                actor_type="tool",
+                actor_id=name,
+                identity=("before_tool", pending.occurrence_identity),
+            )
+
+    async def run_async(self, state: Any) -> None:
+        self.run(state)
+
+
+class _AfterToolHook:
+    allowed_hook_points = frozenset({"after_tool"})
+
+    def __init__(self, owner: "LoopGridHaystack", decision_id: str) -> None:
+        self._owner = owner
+        self._decision_id = decision_id
+
+    def run(self, state: Any) -> None:
+        pending_batch = self._owner._take_pending(state, self._decision_id)
+        if pending_batch is None or not pending_batch.calls:
+            return
+
+        # Only consider results produced by this tool step. This prevents a legitimate later
+        # invocation that reuses the same tool-call ID/arguments from binding to an older result.
+        results = _tool_results_from_state(state)[pending_batch.result_offset :]
+        by_identity: dict[str, list[Any]] = {}
+        for result in results:
+            origin = getattr(result, "origin", None)
+            if origin is not None:
+                by_identity.setdefault(_call_identity(origin), []).append(result)
+
+        for pending in pending_batch.calls:
+            call = pending.call
+            matching_results = by_identity.get(pending.call_identity) or []
+            result = matching_results.pop(0) if matching_results else None
+            if result is None:
+                self._owner._record_error(
+                    "after_tool_missing_result",
+                    self._decision_id,
+                    RuntimeError(f"No Haystack tool result found for {pending.call_identity}"),
+                )
+                continue
+
+            origin = getattr(result, "origin", None) or call
+            name = str(getattr(origin, "tool_name", None) or getattr(call, "tool_name", None) or "tool")
+            raw_result = getattr(result, "result", None)
+            failed = bool(getattr(result, "error", False))
+            payload: dict[str, Any] = {
+                "framework": _FRAMEWORK,
+                "haystack_target": HAYSTACK_TARGET_VERSION,
+                "tool": name,
+                "call_id": getattr(origin, "id", None),
+                "agent_step": pending.step_identity,
+                "tool_ordinal": pending.ordinal,
+                "status": "failed" if failed else "completed",
+                "error": failed,
+                "output_commitment": _commitment(raw_result),
+                "source": "agent.after_tool",
+            }
+            if self._owner.capture_content:
+                payload["output"] = raw_result
+            self._owner._transport.enqueue(
+                self._decision_id,
+                "tool_executed",
+                payload,
+                actor_type="tool",
+                actor_id=name,
+                identity=("after_tool", pending.occurrence_identity),
+            )
+
+    async def run_async(self, state: Any) -> None:
+        self.run(state)
+
+
+class _AfterRunHook:
+    allowed_hook_points = frozenset({"after_run"})
+
+    def __init__(self, owner: "LoopGridHaystack", decision_id: str) -> None:
+        self._owner = owner
+        self._decision_id = decision_id
+
+    def run(self, state: Any) -> None:
+        self._owner._leave_run(state, self._decision_id)
+
+    async def run_async(self, state: Any) -> None:
+        self.run(state)
+
+
+class LoopGridHaystack:
+    """LoopGrid evidence integration for Haystack 3.3 Agents.
+
+    Native surfaces:
+      * Haystack ``Tracer`` / ``Span`` for model-completion evidence.
+      * Agent ``before_tool`` / ``after_tool`` hooks for actual tool execution boundaries.
+      * Explicit application calls for policy, human review, and observed downstream outcome.
+    """
+
+    def __init__(
+        self,
+        *,
+        client: LoopGrid | None = None,
+        base_url: str = "http://127.0.0.1:8000",
+        api_key: str | None = None,
+        workspace_id: str = "default",
+        agent_id: str = "haystack-agent",
+        capture_content: bool = False,
+        default_model_name: str | None = None,
+        default_model_provider: str | None = None,
+    ) -> None:
+        _require_nonempty("agent_id", agent_id)
+        self.client = client or LoopGrid(base_url=base_url, api_key=api_key, workspace_id=workspace_id)
+        self.workspace_id = workspace_id
+        self.agent_id = agent_id
+        self.capture_content = bool(capture_content)
+        self.default_model_name = default_model_name
+        self.default_model_provider = default_model_provider
+
+        self._transport = _EventTransport(self)
+        self._errors: list[dict[str, Any]] = []
+        self._errors_lock = threading.Lock()
+        self._state_lock = threading.RLock()
+        self._pending_by_state: dict[int, _PendingToolBatch] = {}
+        self._previous_context_by_state: dict[int, tuple[str, ...]] = {}
+        self._decision_models: dict[str, dict[str, Any]] = {}
+        self._span_seq: dict[str, int] = {}
+        self._span_seq_lock = threading.Lock()
+        self._decision_stack: contextvars.ContextVar[tuple[str, ...]] = contextvars.ContextVar(
+            "loopgrid_haystack_decision_stack", default=()
+        )
+        self._installed_tracer: LoopGridTracer | None = None
+
+    @property
+    def current_decision_id(self) -> str | None:
+        stack = self._decision_stack.get()
+        return stack[-1] if stack else None
+
+    def _next_span_sequence(self, decision_id: str | None) -> int | None:
+        if not decision_id:
+            return None
+        with self._span_seq_lock:
+            next_value = self._span_seq.get(decision_id, 0) + 1
+            self._span_seq[decision_id] = next_value
+            return next_value
+
+    def make_tracer(self, *, delegate: Tracer | None = None) -> LoopGridTracer:
+        return LoopGridTracer(self, delegate=delegate)
+
+    def enable_tracing(self, *, delegate: Tracer | None = None) -> LoopGridTracer:
+        """Explicitly enable LoopGrid as Haystack's process-level tracer.
+
+        Haystack exposes a process-level tracing facade. The public ``haystack.tracing.tracer``
+        object must not be captured and then used as a delegate: that facade routes to whichever
+        tracer is currently enabled, so delegating to it after installing LoopGrid can recurse
+        back into LoopGrid.
+
+        To compose LoopGrid with another tracing backend, keep the concrete backend tracer
+        instance and pass it explicitly as ``delegate``. Constructing ``LoopGridHaystack`` alone
+        never changes global Haystack tracing.
+        """
+
+        if delegate is not None and (delegate is self._installed_tracer or isinstance(delegate, LoopGridTracer)):
+            raise ValueError("delegate must be a concrete non-LoopGrid Haystack tracer")
+        tracer = self.make_tracer(delegate=delegate)
+        tracing.enable_tracing(tracer)
+        self._installed_tracer = tracer
+        return tracer
+
+    def agent_hooks(
+        self,
+        decision_id: str,
+        *,
+        before_tool_prefix: Sequence[Any] = (),
+        after_tool_suffix: Sequence[Any] = (),
+    ) -> dict[str, list[Any]]:
+        """Return decision-scoped native Haystack Agent hooks.
+
+        Put ``ConfirmationHook`` (and any other hook that may remove/modify tool calls) in
+        ``before_tool_prefix``. LoopGrid's request hook is appended after those controls and
+        therefore records only the calls that will actually proceed to Haystack execution.
+        """
+
+        _require_nonempty("decision_id", decision_id)
+        return {
+            "before_run": [_BeforeRunHook(self, decision_id)],
+            "before_tool": [*list(before_tool_prefix), _BeforeToolHook(self, decision_id)],
+            "after_tool": [_AfterToolHook(self, decision_id), *list(after_tool_suffix)],
+            "after_run": [_AfterRunHook(self, decision_id)],
+        }
+
+    @contextlib.contextmanager
+    def decision_context(self, decision_id: str) -> Iterator[None]:
+        """Exception-safe explicit correlation scope for advanced/custom run loops.
+
+        Normal ``Agent`` usage gets correlation from ``before_run``/``after_run`` hooks. This
+        context manager is useful when application code needs a guaranteed cleanup boundary even
+        if a generator/tool raises before ``after_run`` can execute.
+        """
+
+        _require_nonempty("decision_id", decision_id)
+        previous = self._decision_stack.get()
+        token = self._decision_stack.set((*previous, decision_id))
+        try:
+            yield
+        finally:
+            self._decision_stack.reset(token)
+
+    def start_decision(
+        self,
+        *,
+        decision_type: str,
+        agent: Mapping[str, Any],
+        authority: Mapping[str, Any],
+        model: Mapping[str, Any],
+        context: Mapping[str, Any],
+        proposed_action: Mapping[str, Any],
+        policy: Mapping[str, Any] | None = None,
+        metadata: Mapping[str, Any] | None = None,
+        privacy_mode: str = "redacted",
+        idempotency_key: str | None = None,
+    ) -> dict[str, Any]:
+        _require_nonempty("decision_type", decision_type)
+        _require_nonempty("agent.id", dict(agent).get("id"))
+        _require_nonempty("authority", authority)
+        _require_nonempty("model.name", dict(model).get("name"))
+        _require_nonempty("context", context)
+        _require_nonempty("proposed_action", proposed_action)
+
+        md = dict(metadata or {})
+        md.update(
+            {
+                "framework": _FRAMEWORK,
+                "integration": "loopgrid-haystack",
+                "integration_version": LOOPGRID_HAYSTACK_VERSION,
+                "haystack_target": HAYSTACK_TARGET_VERSION,
+            }
+        )
+        result = self.client.record_decision(
+            decision_type=decision_type,
+            agent=dict(agent),
+            authority=dict(authority),
+            model=dict(model),
+            context=dict(context),
+            proposed_action=dict(proposed_action),
+            metadata=md,
+            privacy_mode=privacy_mode,
+            idempotency_key=idempotency_key,
+        )
+        decision_id = result["decision_id"]
+        with self._state_lock:
+            self._decision_models[decision_id] = dict(model)
+        if policy is not None:
+            self.record_policy(decision_id, policy)
+        return result
+
+    def record_policy(self, decision_id: str, policy: Mapping[str, Any]) -> Any:
+        p = dict(policy)
+        decision = p.get("decision")
+        if decision not in {"auto_allowed", "human_approval_required", "blocked", "block"}:
+            raise ValueError(
+                "policy.decision must be auto_allowed, human_approval_required, blocked, or block"
+            )
+        if not (p.get("policy_id") or p.get("version")):
+            raise ValueError("policy provenance requires policy_id or version")
+        return self.client.policy_evaluated(
+            decision_id,
+            p,
+            actor_id=str(p.get("policy_id") or "application-policy"),
+        )
+
+    def record_human_review(
+        self,
+        decision_id: str,
+        *,
+        reviewer: str,
+        approved: bool,
+        reason: str = "",
+    ) -> Any:
+        _require_nonempty("reviewer", reviewer)
+        # Explicit application facts are written synchronously by the LoopGrid SDK. Drain
+        # framework-generated evidence first so a slow transport cannot invert the real lifecycle
+        # (for example model_completed arriving after human_approved).
+        self.flush()
+        if approved:
+            return self.client.human_approved(decision_id, reviewer, reason)
+        return self.client.human_rejected(decision_id, reviewer, reason)
+
+    def record_outcome(
+        self,
+        decision_id: str,
+        outcome: Mapping[str, Any],
+        *,
+        observer: str = "application",
+    ) -> Any:
+        _require_nonempty("outcome", outcome)
+        # Preserve ordering with queued model/tool evidence before recording the authoritative
+        # downstream observation.
+        self.flush()
+        return self.client.outcome_observed(decision_id, dict(outcome), actor_id=observer)
+
+    def flush(self) -> None:
+        self._transport.flush()
+
+    def shutdown(self) -> None:
+        self._transport.shutdown()
+
+    def _enter_run(self, state: Any, decision_id: str) -> None:
+        state_id = id(state)
+        previous = self._decision_stack.get()
+        with self._state_lock:
+            self._previous_context_by_state[state_id] = previous
+        self._decision_stack.set((*previous, decision_id))
+
+    def _leave_run(self, state: Any, decision_id: str) -> None:
+        state_id = id(state)
+        with self._state_lock:
+            previous = self._previous_context_by_state.pop(state_id, None)
+            self._pending_by_state.pop(state_id, None)
+        if previous is not None:
+            self._decision_stack.set(previous)
+        else:
+            # Fail closed on correlation rather than retaining a stale decision.
+            self._decision_stack.set(())
+
+    def _set_pending(
+        self,
+        state: Any,
+        decision_id: str,
+        calls: list[_PendingToolCall],
+    ) -> None:
+        with self._state_lock:
+            self._pending_by_state[id(state)] = _PendingToolBatch(
+                decision_id=decision_id,
+                calls=tuple(calls),
+                result_offset=len(_tool_results_from_state(state)),
+            )
+
+    def _take_pending(self, state: Any, decision_id: str) -> _PendingToolBatch | None:
+        with self._state_lock:
+            stored = self._pending_by_state.pop(id(state), None)
+        if stored is None:
+            return None
+        if stored.decision_id != decision_id:
+            self._record_error(
+                "pending_decision_mismatch",
+                decision_id,
+                RuntimeError(f"pending tool calls belong to {stored.decision_id}"),
+            )
+            return None
+        return stored
+
+    def _finish_span(self, span: LoopGridSpan, error: BaseException | None) -> None:
+        # Haystack 3.x emits one LLM child span per Agent step. We intentionally ignore
+        # tool spans because hook ordering is the authoritative execution boundary.
+        if span.decision_id is None:
+            return
+        if span.operation_name != "haystack.agent.step.llm":
+            return
+
+        with self._state_lock:
+            model = dict(self._decision_models.get(span.decision_id, {}))
+        model_name = _model_name_from_tags(span.tags, model.get("name") or self.default_model_name)
+        provider = model.get("provider") or self.default_model_provider
+        payload: dict[str, Any] = {
+            "framework": _FRAMEWORK,
+            "haystack_target": HAYSTACK_TARGET_VERSION,
+            "operation": span.operation_name,
+            "parent_operation": span.parent_operation,
+            "provider": provider,
+            "model": model_name,
+            "status": "failed" if error is not None else "completed",
+            "trace_tags_commitment": _commitment(span.tags),
+            "source": "haystack.tracer",
+        }
+        if error is not None:
+            payload["error_type"] = type(error).__name__
+        if self.capture_content:
+            payload["trace_tags"] = span.tags
+
+        self._transport.enqueue(
+            span.decision_id,
+            "model_completed",
+            payload,
+            actor_type="agent",
+            actor_id=self.agent_id,
+            identity=("llm_span", span.sequence, _commitment(span.tags)["sha256"]),
+        )
+
+    def _record_error(self, stage: str, decision_id: str | None, exc: BaseException) -> None:
+        with self._errors_lock:
+            self._errors.append(
+                {
+                    "stage": stage,
+                    "decision_id": decision_id,
+                    "error_type": type(exc).__name__,
+                    # Do not retain arbitrary exception text; it can contain user data/secrets.
+                    "message": "LoopGrid Haystack integration error",
+                }
+            )
+
+    @property
+    def errors(self) -> tuple[dict[str, Any], ...]:
+        with self._errors_lock:
+            return tuple(dict(x) for x in self._errors)
+
+    def assert_healthy(self) -> None:
+        errors = self.errors
+        if errors:
+            first = errors[0]
+            raise RuntimeError(
+                f"LoopGrid Haystack transport/integration failed ({len(errors)} error(s)); "
+                f"first={first['stage']}:{first['error_type']}"
+            )

loopgrid_haystack-0.1.0.dist-info/METADATA

@@ -0,0 +1,241 @@
+Metadata-Version: 2.4
+Name: loopgrid-haystack
+Version: 0.1.0
+Summary: Signed, tamper-evident decision evidence for Haystack Agents using LoopGrid
+Author: LoopGrid
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://loopgrid.io/integrations/haystack/
+Project-URL: Repository, https://github.com/loopgridio/loopgrid-haystack
+Project-URL: Issues, https://github.com/loopgridio/loopgrid-haystack/issues
+Keywords: haystack,haystack-ai,ai-agents,agent-governance,decision-evidence,audit-trail,provenance,tamper-evident,loopgrid
+Classifier: Development Status :: 4 - Beta
+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: Programming Language :: Python :: 3.14
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: loopgrid<0.9,>=0.8.0
+Requires-Dist: haystack-ai<3.4,>=3.3.0
+Provides-Extra: dev
+Requires-Dist: pytest>=8; extra == "dev"
+Requires-Dist: pytest-asyncio>=0.24; extra == "dev"
+Requires-Dist: build>=1.2; extra == "dev"
+Requires-Dist: twine>=6; extra == "dev"
+Dynamic: license-file
+
+# LoopGrid for Haystack
+
+Native Haystack 3.3 Agent hooks + tracing integration for LoopGrid, the evidence plane for AI agents.
+
+`loopgrid-haystack` maps Haystack's real Agent lifecycle into LoopGrid decision evidence without changing the LoopGrid Core contract.
+
+## Evidence contract
+
+```text
+application starts consequential decision
+        ↓
+Haystack Agent LLM span completes
+        ↓
+LoopGrid model_completed
+        ↓
+explicit application policy / human review
+        ↓
+Haystack ConfirmationHook / other controls
+        ↓
+LoopGrid before_tool observes the surviving tool call
+        ↓
+LoopGrid tool_requested
+        ↓
+actual Haystack tool execution
+        ↓
+Haystack after_tool
+        ↓
+LoopGrid tool_executed
+        ↓
+application explicitly observes downstream outcome
+        ↓
+LoopGrid outcome_observed
+        ↓
+evidence_complete → cryptographic verification
+```
+
+LoopGrid **does not** execute tools, invent delegated authority, invent policy, infer a reviewer, treat confirmation as authorization by itself, or infer a real-world outcome from a tool return.
+
+## Compatibility
+
+- Python 3.10–3.14
+- `haystack-ai` 3.3.x
+- `loopgrid` 0.8.x
+- LoopGrid Core validation target: `0.8.1-design-partner`
+
+## Install
+
+```bash
+pip install loopgrid-haystack
+```
+
+## Quickstart
+
+```python
+from haystack.components.agents import Agent
+from haystack.dataclasses import ChatMessage
+from loopgrid_haystack import LoopGridHaystack
+
+loopgrid = LoopGridHaystack(
+    base_url="http://127.0.0.1:8000",
+    workspace_id="default",
+    agent_id="support-agent",
+)
+
+# Explicit: Haystack never enables tracing automatically. If the application already owns
+# a concrete Haystack tracer backend, pass that backend explicitly as delegate=...
+loopgrid.enable_tracing()
+
+decision = loopgrid.start_decision(
+    decision_type="customer_refund",
+    agent={"id": "support-agent", "version": "1"},
+    authority={"acting_for": "Example Store", "scope": ["refund:create"], "limit_usd": 100},
+    model={"provider": "openai", "name": "gpt-5"},
+    context={"prompt_version": "support-v1"},
+    proposed_action={"tool": "refund.create", "amount": 25, "currency": "USD"},
+    policy={
+        "policy_id": "refund-policy",
+        "version": "1",
+        "decision": "auto_allowed",
+        "reason": "Within delegated threshold",
+    },
+)
+
+agent = Agent(
+    chat_generator=...,
+    tools=[...],
+    hooks=loopgrid.agent_hooks(decision["decision_id"]),
+)
+
+result = agent.run(messages=[ChatMessage.from_user("Handle this duplicate charge.")])
+loopgrid.flush()
+loopgrid.assert_healthy()
+
+# Only after the application observes the authoritative downstream business result:
+loopgrid.record_outcome(
+    decision["decision_id"],
+    {"status": "succeeded", "external_reference": "refund_123"},
+    observer="billing-webhook",
+)
+```
+
+## Human approval / ConfirmationHook
+
+Haystack's `ConfirmationHook` remains the execution-control mechanism. LoopGrid records evidence about an application-authenticated reviewer; it does not replace Haystack confirmation or application authorization.
+
+**Ordering is critical:** pass confirmation/modification hooks through `before_tool_prefix`. LoopGrid's request hook is appended after them, so a rejected call cannot become `tool_requested` evidence and a modified call is recorded with the final model-visible parameters that survive confirmation.
+
+```python
+confirmation_hook = ConfirmationHook(confirmation_strategies={...})
+
+agent = Agent(
+    chat_generator=...,
+    tools=[...],
+    hooks=loopgrid.agent_hooks(
+        decision_id,
+        before_tool_prefix=[confirmation_hook],
+    ),
+)
+```
+
+Your authenticated approval UI/strategy should explicitly record the reviewer when the human makes the decision:
+
+```python
+loopgrid.record_human_review(
+    decision_id,
+    reviewer="reviewer@example.com",
+    approved=True,
+    reason="Reviewed refund evidence",
+)
+```
+
+LoopGrid never derives reviewer identity merely from the fact that a Haystack tool call survived `ConfirmationHook`.
+
+`record_human_review()` drains earlier queued framework evidence before writing the explicit
+review event. This preserves the real lifecycle ordering even when the LoopGrid transport is
+temporarily slower than the application thread.
+
+## Native integration surfaces
+
+The integration uses two different Haystack-native boundaries for two different facts:
+
+- `Tracer` / `Span`: `haystack.agent.step.llm` completion → `model_completed`.
+- Agent `before_tool` / `after_tool`: actual surviving tool request and resulting tool message → `tool_requested` / `tool_executed`.
+
+Haystack tool spans are deliberately ignored for action evidence. The hooks are authoritative because they run after confirmation/modification and around actual Agent-owned tool execution.
+
+## Existing Haystack tracing backends
+
+Haystack exposes one process-level tracing facade. `enable_tracing()` is explicit:
+
+```python
+loopgrid.enable_tracing()
+```
+
+Do **not** capture `haystack.tracing.tracer` and pass that facade back as a delegate. The facade routes to the currently enabled tracer; after LoopGrid is installed that would route back into LoopGrid and recurse.
+
+If your application already owns a concrete Haystack tracer instance, compose it explicitly:
+
+```python
+existing_backend = MyConcreteHaystackTracer(...)
+loopgrid.enable_tracing(delegate=existing_backend)
+```
+
+LoopGrid delegates spans directly to that concrete backend while recording its own evidence. Constructing `LoopGridHaystack` by itself never changes global Haystack tracing.
+
+## Privacy
+
+LoopGrid defaults to `capture_content=False`.
+
+Model span tags and tool inputs/outputs are represented by SHA-256 commitments rather than raw content. Haystack itself also disables content tracing by default. Do not enable `HAYSTACK_CONTENT_TRACING_ENABLED=true` unless your application intentionally permits content export.
+
+If you explicitly set `capture_content=True`, LoopGrid will include the tool input/result and captured Haystack trace tags in evidence payloads.
+
+## Failure behavior
+
+Tracer/hook transport is queued and does not raise into the Haystack Agent execution path. Failures are retained and surfaced explicitly:
+
+```python
+loopgrid.flush()
+loopgrid.assert_healthy()
+```
+
+A successful Agent run is not evidence that LoopGrid transport succeeded unless `assert_healthy()` passes.
+
+## Pre-release validation gates
+
+The release candidate is **not a public release**. Required before `v0.1.0`:
+
+- exact `haystack-ai==3.3.0` runtime install
+- semantic/unit tests
+- real deterministic Haystack Agent runtime test
+- real LoopGrid Core E2E
+- real native `ConfirmationHook` approval ordering E2E
+- rejection and modification regression tests
+- parallel tool-call correlation test
+- repeated-call idempotency across later Agent steps, including missing/reused call IDs
+- queued-event ordering before explicit human review/outcome writes
+- real `Agent.run_async()` runtime regression
+- privacy regression
+- concurrency isolation
+- transport-failure observability
+- `evidence_complete`, applicable coverage 100%, `verify.valid=true`, failures `[]`
+- wheel/sdist build + `twine check`
+- fresh wheel install/import test
+- GitHub CI, exact tested commit tag/release
+- PyPI Trusted Publishing + fresh public install
+- website/docs after public install
+- official Haystack Integrations PR after all release gates pass
+
+## License
+
+Apache-2.0.

loopgrid_haystack-0.1.0.dist-info/RECORD

@@ -0,0 +1,7 @@
+loopgrid_haystack/__init__.py,sha256=DxYy48Ug7PAXamgW1hZhREBDpayyKKq7jP2Bm8S4hxk,294
+loopgrid_haystack/integration.py,sha256=xlch-WvIYCb75bD7L2YAgCzxEqoPWEDbp3XoTr-HfNA,29116
+loopgrid_haystack-0.1.0.dist-info/licenses/LICENSE,sha256=z8d0m5b2O9McPEK1xHG_dWgUBT6EfBDz6wA0F7xSPTA,11358
+loopgrid_haystack-0.1.0.dist-info/METADATA,sha256=5-Dg63JONlkoSmCu6PRO43CsvQ4oKoEMik0uIrHZjT0,8671
+loopgrid_haystack-0.1.0.dist-info/WHEEL,sha256=YVMoNqKzERt-wjUZwJ33xBGAwnFl-4cqbYkTtWa4itE,91
+loopgrid_haystack-0.1.0.dist-info/top_level.txt,sha256=qqcKNwZgBH-9Y1Swx6HyaV7tDQvdfPh1OGF8xg3u1dI,18
+loopgrid_haystack-0.1.0.dist-info/RECORD,,

loopgrid_haystack-0.1.0.dist-info/WHEEL

@@ -0,0 +1,5 @@
+Wheel-Version: 1.0
+Generator: setuptools (84.0.0)
+Root-Is-Purelib: true
+Tag: py3-none-any
+

loopgrid_haystack-0.1.0.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.

loopgrid_haystack-0.1.0.dist-info/top_level.txt

@@ -0,0 +1 @@
+loopgrid_haystack