hermes-agent/plugins/memory/honcho/session.py

"""Honcho-based session management for conversation history."""

from __future__ import annotations

import hashlib
import queue
import re
import logging
import threading
from dataclasses import dataclass, field
from datetime import datetime
from typing import Any, Callable, TYPE_CHECKING

from plugins.memory.honcho.client import get_honcho_client
from plugins.memory.honcho.oauth import redact_tokens as _redact_tokens

if TYPE_CHECKING:
    from honcho import Honcho

logger = logging.getLogger(__name__)

# Sentinel to signal the async writer thread to shut down
_ASYNC_SHUTDOWN = object()
_PEER_ID_HASH_LEN = 8
_PEER_ID_HASH_ESCALATION_LENGTHS = (_PEER_ID_HASH_LEN, 12, 16, 24, 32, 64)


class HonchoAuthError(RuntimeError):
    """Auth failure that survived a forced refresh and one retry.

    Raised, not swallowed, so callers can tell a rejected credential from an empty result.
    """


# Matched narrowly: a false positive spends a token rotation, and a lost rotation revokes the grant.
_AUTH_ERROR_MARKERS = (
    "invalid or expired access token",
    "authentication failed",
    "unauthorized",
)

# A 401 in text counts only with HTTP context ("HTTP 401", "status 401"), never as a bare number.
_HTTP_401_RE = re.compile(r"\b(?:http|status(?:[ _]code)?\s*[:=]?)\s*401\b")


def _is_auth_error(exc: BaseException) -> bool:
    status = getattr(exc, "status_code", None) or getattr(exc, "status", None)
    if status == 401:
        return True
    # The transport reported a concrete non-auth status; trust it over text.
    if isinstance(status, int) and status not in (0, 401):
        return False
    text = str(exc).lower()
    if _HTTP_401_RE.search(text):
        return True
    return any(marker in text for marker in _AUTH_ERROR_MARKERS)


def _auth_error_message(exc: BaseException) -> str:
    return (
        "Honcho rejected our credentials and a forced token refresh did not "
        f"recover: {_redact_tokens(str(exc))}. "
        "Re-authenticate with 'hermes honcho setup'."
    )


_REAUTH_REQUIRED_MESSAGE = (
    "Honcho OAuth grant is revoked and cannot be refreshed; "
    "re-authenticate with 'hermes honcho setup'."
)


@dataclass
class HonchoSession:
    """
    A conversation session backed by Honcho.

    Provides a local message cache that syncs to Honcho's
    AI-native memory system for user modeling.
    """

    key: str  # channel:chat_id
    user_peer_id: str  # Honcho peer ID for the user
    assistant_peer_id: str  # Honcho peer ID for the assistant
    honcho_session_id: str  # Honcho session ID
    messages: list[dict[str, Any]] = field(default_factory=list)
    created_at: datetime = field(default_factory=datetime.now)
    updated_at: datetime = field(default_factory=datetime.now)
    metadata: dict[str, Any] = field(default_factory=dict)

    def add_message(self, role: str, content: str, **kwargs: Any) -> None:
        """Add a message to the local cache."""
        msg = {
            "role": role,
            "content": content,
            "timestamp": datetime.now().isoformat(),
            **kwargs,
        }
        self.messages.append(msg)
        self.updated_at = datetime.now()

    def get_history(self, max_messages: int = 50) -> list[dict[str, Any]]:
        """Get message history for LLM context."""
        recent = (
            self.messages[-max_messages:]
            if len(self.messages) > max_messages
            else self.messages
        )
        return [{"role": m["role"], "content": m["content"]} for m in recent]

    def clear(self) -> None:
        """Clear all messages in the session."""
        self.messages = []
        self.updated_at = datetime.now()


class HonchoSessionManager:
    """
    Manages conversation sessions using Honcho.

    Runs alongside hermes' existing SQLite state and file-based memory,
    adding persistent cross-session user modeling via Honcho's AI-native memory.
    """

    def __init__(
        self,
        honcho: Honcho | None = None,
        context_tokens: int | None = None,
        config: Any | None = None,
        runtime_user_peer_name: str | None = None,
        runtime_user_peer_name_alt: str | None = None,
    ):
        """
        Initialize the session manager.

        Args:
            honcho: Optional Honcho client. If not provided, uses the singleton.
            context_tokens: Max tokens for context() calls (None = Honcho default).
            config: HonchoClientConfig from global config (provides peer_name, ai_peer,
                    write_frequency, observation, etc.).
            runtime_user_peer_name: Gateway user identity for per-user memory scoping.
            runtime_user_peer_name_alt: Optional stable alternate gateway identity.
        """
        self._honcho = honcho
        self._context_tokens = context_tokens
        self._config = config
        self._runtime_user_peer_name = runtime_user_peer_name
        self._runtime_user_peer_name_alt = runtime_user_peer_name_alt
        self._cache: dict[str, HonchoSession] = {}
        self._cache_lock = threading.RLock()
        self._peers_cache: dict[str, Any] = {}
        self._sessions_cache: dict[str, Any] = {}
        # Bumped (under _cache_lock) whenever _force_reauth rebuilds the client.
        # In-flight resolvers compare it around their SDK fetch so an object
        # bound to the discarded client is never stored into the fresh cache.
        self._client_generation = 0

        # Set when a call still fails auth after a forced token refresh; cleared on the next success.
        self._auth_failure: str | None = None
        self._auth_notice_emitted = False

        # Write frequency state
        write_frequency = (config.write_frequency if config else "async")
        self._write_frequency = write_frequency
        self._turn_counter: int = 0

        # Prefetch cache: session_key → last context result (consumed once per turn).
        # Dialectic results are cached on the plugin side (HonchoMemoryProvider
        # ._prefetch_result) so session-start prewarm and turn-driven fires share
        # one source of truth; see __init__.py _do_session_init for the prewarm.
        self._context_cache: dict[str, dict] = {}
        self._prefetch_cache_lock = threading.Lock()
        self._dialectic_reasoning_level: str = (
            config.dialectic_reasoning_level if config else "low"
        )
        self._dialectic_dynamic: bool = (
            config.dialectic_dynamic if config else True
        )
        self._dialectic_max_chars: int = (
            config.dialectic_max_chars if config else 600
        )
        self._observation_mode: str = (
            config.observation_mode if config else "directional"
        )
        # Per-peer observation booleans (granular, from config)
        self._user_observe_me: bool = config.user_observe_me if config else True
        self._user_observe_others: bool = config.user_observe_others if config else True
        self._ai_observe_me: bool = config.ai_observe_me if config else True
        self._ai_observe_others: bool = config.ai_observe_others if config else True
        self._message_max_chars: int = (
            config.message_max_chars if config else 25000
        )
        self._dialectic_max_input_chars: int = (
            config.dialectic_max_input_chars if config else 10000
        )

        # Async write queue — the writer thread starts lazily on first enqueue
        # (see _ensure_async_writer). Constructing a manager must not spawn
        # background work or touch the network: unit tests build managers with
        # mocked clients, and an eagerly-started writer raced ahead of the mock
        # and wrote test messages to a live local Honcho.
        self._async_queue: queue.Queue | None = None
        self._async_thread: threading.Thread | None = None
        self._async_thread_lock = threading.Lock()
        if write_frequency == "async":
            self._async_queue = queue.Queue()

    @property
    def honcho(self) -> Honcho:
        """Get the Honcho client, refreshing a near-expiry OAuth token in place.

        Routes every access through ``get_honcho_client`` (which returns the same
        cached singleton) so a long session can't outlive its 1h access token.
        """
        self._honcho = get_honcho_client()
        return self._honcho

    def _record_auth_failure(self, exc: BaseException) -> None:
        detail = _redact_tokens(str(exc))
        if self._auth_failure is None:
            logger.error(
                "Honcho authentication failed and token refresh did not recover; "
                "memory sync and recall are paused until the user re-authenticates: %s",
                detail,
            )
        self._auth_failure = detail

    def _clear_auth_failure(self) -> None:
        if self._auth_failure is not None:
            logger.info("Honcho authentication recovered; memory sync and recall resumed")
            self._auth_failure = None
            self._auth_notice_emitted = False

    def pop_auth_notice(self) -> str | None:
        """Return the pending auth-failure message once; later calls return None."""
        if self._auth_failure is None or self._auth_notice_emitted:
            return None
        self._auth_notice_emitted = True
        return self._auth_failure

    def _reauth_required(self) -> bool:
        """True when the grant is dead and only a new login can fix it.

        Compares the on-disk refresh-token digest, so a new login clears it with no network call.
        """
        try:
            from plugins.memory.honcho import oauth

            # Fast path: no grant in this process is dead — skip config-path
            # resolution entirely (this runs before every SDK call).
            if not oauth.any_dead_grants():
                return False

            from plugins.memory.honcho.client import resolve_config_path

            host = getattr(self._config, "host", "") or ""
            if not host:
                return False
            return oauth.reauth_required(resolve_config_path(), host)
        except Exception:
            return False

    def _force_reauth(self) -> bool:
        """Rotate the token after a 401 and rebind the client.

        False for a static API key, a dead grant, or a failed exchange.
        """
        try:
            from plugins.memory.honcho import oauth
            from plugins.memory.honcho.client import (
                reset_honcho_client,
                resolve_config_path,
            )

            host = getattr(self._config, "host", "") or ""
            if not host:
                return False
            token = oauth.force_refresh_token(resolve_config_path(), host)
            if not token:
                return False
            if not oauth.apply_token_to_client(self.honcho, token):
                # SDK shape changed: rebuild the client and drop objects holding the old transport.
                reset_honcho_client()
                with self._cache_lock:
                    self._client_generation += 1
                    self._peers_cache.clear()
                    self._sessions_cache.clear()
            return True
        except Exception:
            logger.warning("Honcho post-401 token refresh failed", exc_info=True)
            return False

    def _authed_call(self, op_name: str, operation: Callable[[], Any]) -> Any:
        """Run an authenticated SDK operation, forcing one token refresh on a 401.

        ``operation`` must re-resolve peer/session objects itself: a failed
        in-place refresh rebuilds the client, orphaning objects captured earlier.
        """
        if self._reauth_required():
            exc = HonchoAuthError(_REAUTH_REQUIRED_MESSAGE)
            self._record_auth_failure(exc)
            raise exc
        try:
            result = operation()
        except HonchoAuthError:
            raise
        except Exception as e:
            if not _is_auth_error(e):
                raise
            logger.warning(
                "Honcho %s hit an auth error; forcing token refresh and "
                "retrying once: %s", op_name, _redact_tokens(str(e)),
            )
            if not self._force_reauth():
                self._record_auth_failure(e)
                raise HonchoAuthError(_auth_error_message(e)) from e
            try:
                result = operation()
            except Exception as retry_exc:
                if _is_auth_error(retry_exc):
                    self._record_auth_failure(retry_exc)
                    raise HonchoAuthError(_auth_error_message(retry_exc)) from retry_exc
                raise
        self._clear_auth_failure()
        return result

    def _sdk_session(self, session_id: str) -> Any:
        """Get or create the SDK session; a client rebuild clears the cache, so re-fetch."""
        while True:
            with self._cache_lock:
                cached = self._sessions_cache.get(session_id)
                generation = self._client_generation
            if cached is not None:
                return cached
            sdk_session = self.honcho.session(session_id)
            with self._cache_lock:
                if self._client_generation == generation:
                    return self._sessions_cache.setdefault(session_id, sdk_session)
            # The client was rebuilt while we resolved; this object holds the
            # discarded transport. Don't cache it — resolve afresh.

    def _get_or_create_peer(self, peer_id: str) -> Any:
        """Get or create a Honcho peer (one get-or-create API call, then cached)."""
        while True:
            with self._cache_lock:
                if peer_id in self._peers_cache:
                    return self._peers_cache[peer_id]
                generation = self._client_generation

            peer = self._authed_call("peer setup", lambda: self.honcho.peer(peer_id))
            with self._cache_lock:
                if self._client_generation == generation:
                    return self._peers_cache.setdefault(peer_id, peer)
            # Client rebuilt mid-resolve — drop the stale object and retry.

    def _get_or_create_honcho_session(
        self, session_id: str, user_peer: Any, assistant_peer: Any
    ) -> tuple[Any, list]:
        """
        Get or create a Honcho session with peers configured.

        Returns:
            Tuple of (honcho_session, existing_messages).
        """
        with self._cache_lock:
            if session_id in self._sessions_cache:
                logger.debug("Honcho session '%s' retrieved from cache", session_id)
                return self._sessions_cache[session_id], []

        self._authed_call("session setup", lambda: self._sdk_session(session_id))

        # Configure per-peer observation from granular booleans.
        # These map 1:1 to Honcho's SessionPeerConfig toggles.
        auth_dead = False
        try:
            from honcho.session import SessionPeerConfig
            user_config = SessionPeerConfig(
                observe_me=self._user_observe_me,
                observe_others=self._user_observe_others,
            )
            ai_config = SessionPeerConfig(
                observe_me=self._ai_observe_me,
                observe_others=self._ai_observe_others,
            )
            peer_entries = [(user_peer, user_config), (assistant_peer, ai_config)]

            self._authed_call(
                "session peer setup",
                lambda: self._sdk_session(session_id).add_peers(peer_entries),
            )

            # Sync back: server-side config (set via Honcho UI) wins over
            # local defaults. Read the effective config after add_peers.
            # Note: observation booleans are manager-scoped, not per-session.
            # Last session init wins. Fine for CLI; gateway should scope per-session.
            try:
                def _read_server_configs() -> tuple[Any, Any]:
                    sdk_session = self._sdk_session(session_id)
                    return (
                        sdk_session.get_peer_configuration(user_peer),
                        sdk_session.get_peer_configuration(assistant_peer),
                    )

                server_user, server_ai = self._authed_call(
                    "peer configuration read", _read_server_configs
                )
                if server_user.observe_me is not None:
                    self._user_observe_me = server_user.observe_me
                if server_user.observe_others is not None:
                    self._user_observe_others = server_user.observe_others
                if server_ai.observe_me is not None:
                    self._ai_observe_me = server_ai.observe_me
                if server_ai.observe_others is not None:
                    self._ai_observe_others = server_ai.observe_others
                logger.debug(
                    "Honcho observation synced from server: user(me=%s,others=%s) ai(me=%s,others=%s)",
                    self._user_observe_me, self._user_observe_others,
                    self._ai_observe_me, self._ai_observe_others,
                )
            except HonchoAuthError:
                raise
            except Exception as e:
                logger.debug("Honcho get_peer_configuration failed (using local config): %s", e)
        except HonchoAuthError:
            # Already recorded by _authed_call; skip the remaining init calls.
            auth_dead = True
        except Exception as e:
            logger.warning(
                "Honcho session '%s' add_peers failed (non-fatal): %s",
                session_id, e,
            )

        # Load existing messages via context() - single call for messages + metadata
        existing_messages = []
        if not auth_dead:
            try:
                ctx = self._authed_call(
                    "session context load",
                    lambda: self._sdk_session(session_id).context(
                        summary=True, tokens=self._context_tokens
                    ),
                )
                existing_messages = ctx.messages or []

                # Verify chronological ordering
                if existing_messages and len(existing_messages) > 1:
                    timestamps = [m.created_at for m in existing_messages if m.created_at]
                    if timestamps and timestamps != sorted(timestamps):
                        logger.warning(
                            "Honcho messages not chronologically ordered for session '%s', sorting",
                            session_id,
                        )
                        existing_messages = sorted(
                            existing_messages,
                            key=lambda m: m.created_at or datetime.min,
                        )

                if existing_messages:
                    logger.info(
                        "Honcho session '%s' retrieved (%d existing messages)",
                        session_id, len(existing_messages),
                    )
                else:
                    logger.info("Honcho session '%s' created (new)", session_id)
            except HonchoAuthError:
                logger.warning(
                    "Honcho session '%s' loaded without server context: auth failed",
                    session_id,
                )
            except Exception as e:
                logger.warning(
                    "Honcho session '%s' loaded (failed to fetch context: %s)",
                    session_id, e,
                )

        with self._cache_lock:
            honcho_session = self._sessions_cache.get(session_id)
        if honcho_session is None:
            # A mid-init client rebuild dropped the cached session; resolve a fresh one.
            honcho_session = self._authed_call(
                "session setup", lambda: self._sdk_session(session_id)
            )
        return honcho_session, existing_messages

    def _sanitize_id(self, id_str: str) -> str:
        """Sanitize an ID to match Honcho's pattern: ^[a-zA-Z0-9_-]+"""
        return re.sub(r'[^a-zA-Z0-9_-]', '-', id_str)

    def _runtime_user_ids(self) -> list[str]:
        """Return runtime identity candidates in lookup order."""
        candidates: list[str] = []
        for value in (self._runtime_user_peer_name, self._runtime_user_peer_name_alt):
            if value is None:
                continue
            candidate = str(value).strip()
            if candidate and candidate not in candidates:
                candidates.append(candidate)
        return candidates

    def _session_key_fallback_peer_id(self, key: str) -> str:
        parts = key.split(":", 1)
        channel = parts[0] if len(parts) > 1 else "default"
        chat_id = parts[1] if len(parts) > 1 else key
        return self._sanitize_id(f"user-{channel}-{chat_id}")

    def _explicit_user_peer_ids(self) -> set[str]:
        """Return sanitized user peer IDs that came from explicit config."""
        if self._config is None:
            return set()

        explicit_ids: set[str] = set()
        peer_name = getattr(self._config, "peer_name", None)
        if peer_name:
            explicit_ids.add(self._sanitize_id(str(peer_name).strip()))

        aliases = getattr(self._config, "user_peer_aliases", {})
        if isinstance(aliases, dict):
            for alias in aliases.values():
                if isinstance(alias, str) and alias.strip():
                    explicit_ids.add(self._sanitize_id(alias.strip()))

        return explicit_ids

    def _generated_runtime_peer_id(self, prefix: str, runtime_id: str) -> str:
        """Return a stable peer ID for an unknown prefixed runtime user."""
        raw_peer_id = f"{prefix}{runtime_id}"
        sanitized_peer_id = self._sanitize_id(raw_peer_id)
        explicit_ids = self._explicit_user_peer_ids()
        if (
            sanitized_peer_id != raw_peer_id
            or sanitized_peer_id in explicit_ids
        ):
            digest = hashlib.sha256(raw_peer_id.encode("utf-8")).hexdigest()
            for hash_len in _PEER_ID_HASH_ESCALATION_LENGTHS:
                candidate = f"{sanitized_peer_id}-{digest[:hash_len]}"
                if candidate not in explicit_ids:
                    return candidate
            return f"{sanitized_peer_id}-{digest}"
        return sanitized_peer_id

    def _resolve_user_peer_id(self, key: str) -> str:
        """Resolve the Honcho user peer ID for this manager/session."""
        pin_peer_name = (
            self._config is not None
            and bool(getattr(self._config, "peer_name", None))
            and getattr(self._config, "pin_peer_name", False) is True
        )
        if pin_peer_name:
            return self._sanitize_id(self._config.peer_name)

        runtime_ids = self._runtime_user_ids()
        if runtime_ids:
            aliases = getattr(self._config, "user_peer_aliases", {}) if self._config else {}
            if not isinstance(aliases, dict):
                aliases = {}
            for runtime_id in runtime_ids:
                alias = aliases.get(runtime_id)
                if isinstance(alias, str) and alias.strip():
                    return self._sanitize_id(alias.strip())

            primary_runtime_id = runtime_ids[0]
            prefix = getattr(self._config, "runtime_peer_prefix", "") if self._config else ""
            prefix = prefix.strip() if isinstance(prefix, str) else ""
            if prefix:
                return self._generated_runtime_peer_id(prefix, primary_runtime_id)
            return self._sanitize_id(primary_runtime_id)

        if self._config and self._config.peer_name:
            return self._sanitize_id(self._config.peer_name)

        return self._session_key_fallback_peer_id(key)

    def get_or_create(self, key: str) -> HonchoSession:
        """
        Get an existing session or create a new one.

        Args:
            key: Session key (usually channel:chat_id).

        Returns:
            The session.
        """
        with self._cache_lock:
            if key in self._cache:
                logger.debug("Local session cache hit: %s", key)
                return self._cache[key]

        # Determine peer IDs — no lock needed (read-only, no shared state mutation).
        # Gateway sessions normally use the runtime user identity (the
        # platform-native ID: Telegram UID, Discord snowflake, Slack user,
        # etc.) so multi-user bots scope memory per user.  Config can alias
        # known runtime IDs or prefix unknown IDs.  For a single-user
        # deployment, ``pinPeerName`` still pins all runtime identities to
        # ``peerName`` (see #14984).
        user_peer_id = self._resolve_user_peer_id(key)

        assistant_peer_id = self._sanitize_id(
            self._config.ai_peer if self._config else "hermes-assistant"
        )

        # All expensive I/O outside the lock — Honcho's persistence is source of truth
        honcho_session_id = self._sanitize_id(key)
        user_peer = self._get_or_create_peer(user_peer_id)
        assistant_peer = self._get_or_create_peer(assistant_peer_id)
        honcho_session, existing_messages = self._get_or_create_honcho_session(
            honcho_session_id, user_peer, assistant_peer
        )

        local_messages = []
        for msg in existing_messages:
            role = "assistant" if msg.peer_id == assistant_peer_id else "user"
            local_messages.append({
                "role": role,
                "content": msg.content,
                "timestamp": msg.created_at.isoformat() if msg.created_at else "",
                "_synced": True,
            })

        session = HonchoSession(
            key=key,
            user_peer_id=user_peer_id,
            assistant_peer_id=assistant_peer_id,
            honcho_session_id=honcho_session_id,
            messages=local_messages,
        )

        # Write to cache under lock — only one writer wins
        with self._cache_lock:
            self._cache[key] = session
        return session

    def _flush_session(self, session: HonchoSession) -> bool:
        """Internal: write unsynced messages to Honcho synchronously."""
        if not session.messages:
            return True

        new_messages = [m for m in session.messages if not m.get("_synced")]
        if not new_messages:
            return True

        # Resolved inside the operation so a retry after a client rebuild gets fresh objects.
        def _sync_messages() -> int:
            user_peer = self._get_or_create_peer(session.user_peer_id)
            assistant_peer = self._get_or_create_peer(session.assistant_peer_id)
            honcho_session = self._sessions_cache.get(session.honcho_session_id)
            if honcho_session is None:
                honcho_session, _ = self._get_or_create_honcho_session(
                    session.honcho_session_id, user_peer, assistant_peer
                )
            honcho_messages = [
                (user_peer if m["role"] == "user" else assistant_peer).message(m["content"])
                for m in new_messages
            ]
            honcho_session.add_messages(honcho_messages)
            return len(honcho_messages)

        try:
            synced = self._authed_call("message sync", _sync_messages)
            for msg in new_messages:
                msg["_synced"] = True
            logger.debug("Synced %d messages to Honcho for %s", synced, session.key)
            with self._cache_lock:
                self._cache[session.key] = session
            return True
        except Exception as e:
            for msg in new_messages:
                msg["_synced"] = False
            logger.error("Failed to sync messages to Honcho: %s", e)
            with self._cache_lock:
                self._cache[session.key] = session
            return False

    def _async_writer_loop(self) -> None:
        """Background daemon thread: drains the async write queue."""
        while True:
            try:
                item = self._async_queue.get(timeout=5)
                if item is _ASYNC_SHUTDOWN:
                    break

                first_error: Exception | None = None
                try:
                    success = self._flush_session(item)
                except Exception as e:
                    success = False
                    first_error = e

                if success:
                    continue

                if first_error is not None:
                    logger.warning("Honcho async write failed, retrying once: %s", first_error)
                else:
                    logger.warning("Honcho async write failed, retrying once")

                import time as _time
                _time.sleep(2)

                try:
                    retry_success = self._flush_session(item)
                except Exception as e2:
                    logger.error("Honcho async write retry failed, dropping batch: %s", e2)
                    continue

                if not retry_success:
                    logger.error("Honcho async write retry failed, dropping batch")
            except queue.Empty:
                continue
            except Exception as e:
                logger.error("Honcho async writer error: %s", e)

    def save(self, session: HonchoSession) -> None:
        """Save messages to Honcho, respecting write_frequency.

        write_frequency modes:
          "async"   — enqueue for background thread (zero blocking, zero token cost)
          "turn"    — flush synchronously every turn
          "session" — defer until flush_session() is called explicitly
          N (int)   — flush every N turns
        """
        self._turn_counter += 1
        wf = self._write_frequency

        if wf == "async":
            if self._async_queue is not None:
                self._ensure_async_writer()
                self._async_queue.put(session)
        elif wf == "turn":
            self._flush_session(session)
        elif wf == "session":
            # Accumulate; caller must call flush_all() at session end
            pass
        elif isinstance(wf, int) and wf > 0:
            if self._turn_counter % wf == 0:
                self._flush_session(session)

    def flush_all(self) -> None:
        """Flush all pending unsynced messages for all cached sessions.

        Called at session end for "session" write_frequency, or to force
        a sync before process exit regardless of mode.
        """
        with self._cache_lock:
            sessions = list(self._cache.values())
        for session in sessions:
            try:
                self._flush_session(session)
            except Exception as e:
                logger.error("Honcho flush_all error for %s: %s", session.key, e)

        # Drain async queue synchronously if it exists
        if self._async_queue is not None:
            while not self._async_queue.empty():
                try:
                    item = self._async_queue.get_nowait()
                    if item is not _ASYNC_SHUTDOWN:
                        self._flush_session(item)
                except queue.Empty:
                    break

    def _ensure_async_writer(self) -> None:
        """Start the async writer on first enqueue (idempotent, thread-safe)."""
        if self._async_thread is not None and self._async_thread.is_alive():
            return
        with self._async_thread_lock:
            if self._async_thread is None or not self._async_thread.is_alive():
                self._async_thread = threading.Thread(
                    target=self._async_writer_loop,
                    name="honcho-async-writer",
                    daemon=True,
                )
                self._async_thread.start()

    def shutdown(self) -> None:
        """Gracefully shut down the async writer thread."""
        if self._async_queue is not None:
            self.flush_all()
            if self._async_thread is not None and self._async_thread.is_alive():
                self._async_queue.put(_ASYNC_SHUTDOWN)
                self._async_thread.join(timeout=10)

    def delete(self, key: str) -> bool:
        """Delete a session from local cache."""
        with self._cache_lock:
            if key in self._cache:
                del self._cache[key]
                return True
        return False

    def new_session(self, key: str) -> HonchoSession:
        """
        Create a new session, preserving the old one for user modeling.

        Creates a fresh session with a new ID while keeping the old
        session's data in Honcho for continued user modeling.
        """
        import time

        # Hold the reentrant lock across get_or_create so a concurrent caller
        # can't observe the (old-popped, new-not-yet-inserted) gap and create
        # its own session under the raw key.  `_cache_lock` is an RLock so
        # nested reacquisition inside get_or_create is safe.
        with self._cache_lock:
            # Remove old session from caches (but don't delete from Honcho)
            old_session = self._cache.pop(key, None)
            if old_session:
                self._sessions_cache.pop(old_session.honcho_session_id, None)

            # Create new session with timestamp suffix
            timestamp = int(time.time())
            new_key = f"{key}:{timestamp}"

            # get_or_create will create a fresh session
            session = self.get_or_create(new_key)

            # Cache under the original key so callers find it by the expected name
            self._cache[key] = session

        logger.info("Created new session for %s (honcho: %s)", key, session.honcho_session_id)
        return session

    _REASONING_LEVELS = ("minimal", "low", "medium", "high", "max")

    def _default_reasoning_level(self) -> str:
        """Return the configured default reasoning level."""
        return self._dialectic_reasoning_level

    def dialectic_query(
        self, session_key: str, query: str,
        reasoning_level: str | None = None,
        peer: str = "user",
        apply_injection_cap: bool = True,
    ) -> str:
        """
        Query Honcho's dialectic endpoint about a peer.

        Runs an LLM on Honcho's backend against the target peer's full
        representation. Higher latency than context() — callers run this in
        a background thread (see HonchoMemoryProvider) to avoid blocking.

        Args:
            session_key: The session key to query against.
            query: Natural language question.
            reasoning_level: Override the configured default (dialecticReasoningLevel).
                             Only honored when dialecticDynamic is true.
                             If None or dialecticDynamic is false, uses the configured default.
            peer: Which peer to query — "user" (default) or "ai".
            apply_injection_cap: Clip automatic injections to
                ``dialecticMaxChars``. Explicit ``honcho_reasoning`` calls pass
                False because Honcho already bounds their output.

        Returns:
            Honcho's synthesized answer, or empty string on failure.

        Raises:
            HonchoAuthError: the backend rejected our credentials and a forced
                token refresh plus one retry did not recover.
        """
        session = self._cache.get(session_key)
        if not session:
            return ""

        target_peer_id = self._resolve_peer_id(session, peer)
        if target_peer_id is None:
            return ""

        # Guard: truncate query to Honcho's dialectic input limit
        if len(query) > self._dialectic_max_input_chars:
            query = query[:self._dialectic_max_input_chars].rsplit(" ", 1)[0]

        if self._dialectic_dynamic and reasoning_level:
            level = reasoning_level
        else:
            level = self._default_reasoning_level()

        def _chat_once() -> str:
            if self._ai_observe_others:
                # AI peer can observe other peers — use assistant as observer.
                ai_peer_obj = self._get_or_create_peer(session.assistant_peer_id)
                if target_peer_id == session.assistant_peer_id:
                    return ai_peer_obj.chat(query, reasoning_level=level) or ""
                return ai_peer_obj.chat(
                    query,
                    target=target_peer_id,
                    reasoning_level=level,
                ) or ""
            # Without cross-observation, each peer queries its own context.
            target_peer = self._get_or_create_peer(target_peer_id)
            return target_peer.chat(query, reasoning_level=level) or ""

        try:
            result = self._authed_call("dialectic query", _chat_once)
            # Only automatic injection uses the Hermes-side character cap.
            if (
                apply_injection_cap
                and result
                and self._dialectic_max_chars
                and len(result) > self._dialectic_max_chars
            ):
                result = result[:self._dialectic_max_chars].rsplit(" ", 1)[0] + " …"
            return result
        except HonchoAuthError:
            raise
        except Exception as e:
            logger.warning("Honcho dialectic query failed: %s", e)
            return ""

    def prefetch_context(self, session_key: str, user_message: str | None = None) -> None:
        """
        Fire get_prefetch_context in a background thread, caching the result.

        Non-blocking. Consumed next turn via pop_context_result(). This avoids
        a synchronous HTTP round-trip blocking every response.
        """
        def _run():
            result = self.get_prefetch_context(session_key, user_message)
            if result:
                self.set_context_result(session_key, result)

        t = threading.Thread(target=_run, name="honcho-context-prefetch", daemon=True)
        t.start()

    def set_context_result(self, session_key: str, result: dict[str, str]) -> None:
        """Store a prefetched context result in a thread-safe way."""
        if not result:
            return
        with self._prefetch_cache_lock:
            self._context_cache[session_key] = result

    def pop_context_result(self, session_key: str) -> dict[str, str]:
        """
        Return and clear the cached context result for this session.

        Returns empty dict if no result is ready yet (first turn).
        """
        with self._prefetch_cache_lock:
            return self._context_cache.pop(session_key, {})

    def get_prefetch_context(self, session_key: str, user_message: str | None = None) -> dict[str, str]:
        """
        Pre-fetch user and AI peer context from Honcho.

        Fetches peer_representation and peer_card for both peers, plus the
        session summary when available. When user_message is provided, it is
        passed as search_query to the peer context call so Honcho returns
        conclusions relevant to the session topic rather than the full
        observation dump.

        Args:
            session_key: The session key to get context for.
            user_message: Optional first user message used as search_query for
                          topic-relevant context retrieval.

        Returns:
            Dictionary with 'representation', 'card', 'ai_representation',
            'ai_card', and optionally 'summary' keys.
        """
        session = self._cache.get(session_key)
        if not session:
            return {}

        result: dict[str, str] = {}

        # Session summary — provides session-scoped context.
        # Fresh sessions (per-session cold start, or first-ever per-directory)
        # return null summary — the guard below handles that gracefully.
        # Per-directory returning sessions get their accumulated summary.
        try:
            if session.honcho_session_id in self._sessions_cache:
                ctx = self._authed_call(
                    "session summary fetch",
                    lambda: self._sdk_session(session.honcho_session_id).context(summary=True),
                )
                if ctx.summary and getattr(ctx.summary, "content", None):
                    result["summary"] = ctx.summary.content
        except HonchoAuthError:
            # Auth is dead; the pop_auth_notice path tells the model why context is missing.
            return result
        except Exception as e:
            logger.debug("Failed to fetch session summary from Honcho: %s", e)

        try:
            observer_peer_id, target_peer_id = self._resolve_observer_target(session, "user")
            user_ctx = self._fetch_peer_context(observer_peer_id, search_query=user_message or None, target=target_peer_id or session.user_peer_id)
            result["representation"] = user_ctx["representation"]
            result["card"] = "\n".join(user_ctx["card"])
        except HonchoAuthError:
            return result
        except Exception as e:
            logger.warning("Failed to fetch user context from Honcho: %s", e)

        # Also fetch AI peer's own representation so Hermes knows itself.
        try:
            ai_ctx = self._fetch_peer_context(session.assistant_peer_id, target=session.assistant_peer_id)
            result["ai_representation"] = ai_ctx["representation"]
            result["ai_card"] = "\n".join(ai_ctx["card"])
        except HonchoAuthError:
            return result
        except Exception as e:
            logger.debug("Failed to fetch AI peer context from Honcho: %s", e)

        return result

    def migrate_local_history(self, session_key: str, messages: list[dict[str, Any]]) -> bool:
        """
        Upload local session history to Honcho as a file.

        Used when Honcho activates mid-conversation to preserve prior context.

        Args:
            session_key: The session key (e.g., "telegram:123456").
            messages: Local messages (dicts with role, content, timestamp).

        Returns:
            True if upload succeeded, False otherwise.
        """
        session = self._cache.get(session_key)
        if not session:
            logger.warning("No local session cached for '%s', skipping migration", session_key)
            return False

        if session.honcho_session_id not in self._sessions_cache:
            logger.warning("No Honcho session cached for '%s', skipping migration", session_key)
            return False

        content_bytes = self._format_migration_transcript(session_key, messages)
        first_ts = messages[0].get("timestamp") if messages else None

        try:
            def _upload() -> None:
                user_peer = self._get_or_create_peer(session.user_peer_id)
                self._sdk_session(session.honcho_session_id).upload_file(
                    file=("prior_history.txt", content_bytes, "text/plain"),
                    peer=user_peer,
                    metadata={"source": "local_jsonl", "count": len(messages)},
                    created_at=first_ts,
                )

            self._authed_call("history migration upload", _upload)
            logger.info("Migrated %d local messages to Honcho for %s", len(messages), session_key)
            return True
        except Exception as e:
            logger.error("Failed to upload local history to Honcho for %s: %s", session_key, e)
            return False

    @staticmethod
    def _format_migration_transcript(session_key: str, messages: list[dict[str, Any]]) -> bytes:
        """Format local messages as an XML transcript for Honcho file upload."""
        timestamps = [m.get("timestamp", "") for m in messages]
        time_range = f"{timestamps[0]} to {timestamps[-1]}" if timestamps else "unknown"

        lines = [
            "<prior_conversation_history>",
            "<context>",
            "This conversation history occurred BEFORE the Honcho memory system was activated.",
            "These messages are the preceding elements of this conversation session and should",
            "be treated as foundational context for all subsequent interactions. The user and",
            "assistant have already established rapport through these exchanges.",
            "</context>",
            "",
            f'<transcript session_key="{session_key}" message_count="{len(messages)}"',
            f'           time_range="{time_range}">',
            "",
        ]
        for msg in messages:
            ts = msg.get("timestamp", "?")
            role = msg.get("role", "unknown")
            content = msg.get("content") or ""
            lines.append(f"[{ts}] {role}: {content}")

        lines.append("")
        lines.append("</transcript>")
        lines.append("</prior_conversation_history>")

        return "\n".join(lines).encode("utf-8")

    def migrate_memory_files(self, session_key: str, memory_dir: str) -> bool:
        """
        Upload MEMORY.md and USER.md to Honcho as files.

        Used when Honcho activates on an instance that already has locally
        consolidated memory. Backwards compatible -- skips if files don't exist.

        Args:
            session_key: The session key to associate files with.
            memory_dir: Path to the memories directory (~/.hermes/memories/).

        Returns:
            True if at least one file was uploaded, False otherwise.
        """
        from pathlib import Path
        memory_path = Path(memory_dir)

        if not memory_path.exists():
            return False

        session = self._cache.get(session_key)
        if not session:
            logger.warning("No local session cached for '%s', skipping memory migration", session_key)
            return False

        if session.honcho_session_id not in self._sessions_cache:
            logger.warning("No Honcho session cached for '%s', skipping memory migration", session_key)
            return False

        uploaded = False
        files = [
            (
                "MEMORY.md",
                "consolidated_memory.md",
                "Long-term agent notes and preferences",
                session.user_peer_id,
                "user",
            ),
            (
                "USER.md",
                "user_profile.md",
                "User profile and preferences",
                session.user_peer_id,
                "user",
            ),
            (
                "SOUL.md",
                "agent_soul.md",
                "Agent persona and identity configuration",
                session.assistant_peer_id,
                "ai",
            ),
        ]

        for filename, upload_name, description, target_peer_id, target_kind in files:
            filepath = memory_path / filename
            if not filepath.exists():
                continue
            content = filepath.read_text(encoding="utf-8").strip()
            if not content:
                continue

            wrapped = (
                f"<prior_memory_file>\n"
                f"<context>\n"
                f"This file was consolidated from local conversations BEFORE Honcho was activated.\n"
                f"{description}. Treat as foundational context for this user.\n"
                f"</context>\n"
                f"\n"
                f"{content}\n"
                f"</prior_memory_file>\n"
            )

            try:
                def _upload() -> None:
                    self._sdk_session(session.honcho_session_id).upload_file(
                        file=(upload_name, wrapped.encode("utf-8"), "text/plain"),
                        peer=self._get_or_create_peer(target_peer_id),
                        metadata={
                            "source": "local_memory",
                            "original_file": filename,
                            "target_peer": target_kind,
                        },
                    )

                self._authed_call("memory migration upload", _upload)
                logger.info(
                    "Uploaded %s to Honcho for %s (%s peer)",
                    filename,
                    session_key,
                    target_kind,
                )
                uploaded = True
            except HonchoAuthError:
                logger.error("Honcho memory migration stopped after %s: auth failed", filename)
                break
            except Exception as e:
                logger.error("Failed to upload %s to Honcho: %s", filename, e)

        return uploaded

    @staticmethod
    def _normalize_card(card: Any) -> list[str]:
        """Normalize Honcho card payloads into a plain list of strings."""
        if not card:
            return []
        if isinstance(card, list):
            return [str(item) for item in card if item]
        return [str(card)]

    def _fetch_peer_card(self, peer_id: str, *, target: str | None = None) -> list[str]:
        """Fetch a peer card directly from the peer object.

        This avoids relying on session.context(), which can return an empty
        peer_card for per-session messaging sessions even when the peer itself
        has a populated card.
        """
        def _get_card() -> Any:
            peer = self._get_or_create_peer(peer_id)
            getter = getattr(peer, "get_card", None)
            if callable(getter):
                return getter(target=target) if target is not None else getter()
            legacy_getter = getattr(peer, "card", None)
            if callable(legacy_getter):
                return legacy_getter(target=target) if target is not None else legacy_getter()
            return None

        return self._normalize_card(self._authed_call("peer card fetch", _get_card))

    def _fetch_peer_context(
        self,
        peer_id: str,
        search_query: str | None = None,
        *,
        target: str | None = None,
    ) -> dict[str, Any]:
        """Fetch representation + peer card directly from a peer object.

        Raises HonchoAuthError when auth is dead or a 401 survives the forced
        refresh; the non-auth fallback chain would just repeat the failure.
        """
        representation = ""
        card: list[str] = []

        try:
            context_kwargs: dict[str, Any] = {}
            if target is not None:
                context_kwargs["target"] = target
            if search_query is not None:
                context_kwargs["search_query"] = search_query

            def _peer_context() -> Any:
                peer = self._get_or_create_peer(peer_id)
                return peer.context(**context_kwargs) if context_kwargs else peer.context()

            ctx = self._authed_call("peer context fetch", _peer_context)
            representation = (
                getattr(ctx, "representation", None)
                or getattr(ctx, "peer_representation", None)
                or ""
            )
            card = self._normalize_card(getattr(ctx, "peer_card", None))
        except HonchoAuthError:
            raise
        except Exception as e:
            logger.debug("Direct peer.context() failed for '%s': %s", peer_id, e)

        if not representation:
            try:
                def _peer_representation() -> Any:
                    peer = self._get_or_create_peer(peer_id)
                    return peer.representation(target=target) if target is not None else peer.representation()

                representation = self._authed_call(
                    "peer representation fetch", _peer_representation
                ) or ""
            except HonchoAuthError:
                raise
            except Exception as e:
                logger.debug("Direct peer.representation() failed for '%s': %s", peer_id, e)

        if not card:
            try:
                card = self._fetch_peer_card(peer_id, target=target)
            except HonchoAuthError:
                raise
            except Exception as e:
                logger.debug("Direct peer card fetch failed for '%s': %s", peer_id, e)

        return {"representation": representation, "card": card}

    def get_session_context(self, session_key: str, peer: str = "user") -> dict[str, Any]:
        """Fetch full session context from Honcho including summary.

        Uses the session-level context() API which returns summary,
        peer_representation, peer_card, and messages.
        Raises HonchoAuthError so callers can tell rejected credentials from no context.
        """
        session = self._cache.get(session_key)
        if not session:
            return {}

        if session.honcho_session_id not in self._sessions_cache:
            # Fall back to peer-level context, respecting the requested peer
            peer_id = self._resolve_peer_id(session, peer)
            if peer_id is None:
                peer_id = session.user_peer_id
            return self._fetch_peer_context(peer_id, target=peer_id)

        try:
            observer_peer_id, target_peer_id = self._resolve_observer_target(session, peer)
            ctx = self._authed_call(
                "session context fetch",
                lambda: self._sdk_session(session.honcho_session_id).context(
                    summary=True,
                    peer_target=target_peer_id or observer_peer_id,
                    peer_perspective=observer_peer_id,
                ),
            )

            result: dict[str, Any] = {}

            # Summary
            if ctx.summary:
                result["summary"] = ctx.summary.content

            # Peer representation and card
            if ctx.peer_representation:
                result["representation"] = ctx.peer_representation
            if ctx.peer_card:
                result["card"] = "\n".join(ctx.peer_card)

            # Messages (last N for context)
            if ctx.messages:
                recent = ctx.messages[-10:]  # last 10 messages
                result["recent_messages"] = [
                    {"role": getattr(m, "peer_id", "unknown"), "content": (m.content or "")[:500]}
                    for m in recent
                ]

            return result
        except HonchoAuthError:
            raise
        except Exception as e:
            logger.debug("Session context fetch failed: %s", e)
            return {}

    def _resolve_peer_id(self, session: HonchoSession, peer: str | None) -> str:
        """Resolve a peer alias or explicit peer ID to a concrete Honcho peer ID.

        Always returns a non-empty string: either a known peer ID or a
        sanitized version of the caller-supplied alias/ID.
        """
        candidate = (peer or "user").strip()
        if not candidate:
            return session.user_peer_id

        normalized = self._sanitize_id(candidate)
        if normalized == self._sanitize_id("user"):
            return session.user_peer_id
        if normalized == self._sanitize_id("ai"):
            return session.assistant_peer_id

        return normalized

    def _resolve_observer_target(
        self,
        session: HonchoSession,
        peer: str | None,
    ) -> tuple[str, str | None]:
        """Resolve observer and target peer IDs for context/search/profile queries."""
        target_peer_id = self._resolve_peer_id(session, peer)

        if target_peer_id == session.assistant_peer_id:
            return session.assistant_peer_id, session.assistant_peer_id

        if self._ai_observe_others:
            return session.assistant_peer_id, target_peer_id

        return target_peer_id, None

    def get_peer_card(self, session_key: str, peer: str = "user") -> list[str]:
        """
        Fetch a peer card — a curated list of key facts.

        Fast, no LLM reasoning. Returns raw structured facts Honcho has
        inferred about the target peer (name, role, preferences, patterns).
        Empty list if unavailable; raises HonchoAuthError on rejected credentials.
        """
        session = self._cache.get(session_key)
        if not session:
            return []

        try:
            observer_peer_id, target_peer_id = self._resolve_observer_target(session, peer)
            card = self._fetch_peer_card(observer_peer_id, target=target_peer_id)
            if card:
                return card
            # Some backends store cards directly on the target peer, not the
            # observer-target slot. Fall back so honcho_profile still works.
            if target_peer_id:
                return self._fetch_peer_card(target_peer_id)
            return []
        except HonchoAuthError:
            raise
        except Exception as e:
            logger.debug("Failed to fetch peer card from Honcho: %s", e)
            return []

    def search_context(
        self,
        session_key: str,
        query: str,
        max_tokens: int = 800,
        peer: str = "user",
    ) -> str:
        """
        Search raw messages across every session visible from the target
        peer's perspective. Results include all authors and require no LLM
        synthesis.

        Args:
            session_key: Session whose workspace/peer scope to search within.
            query: Search query (hybrid semantic + full-text).
            max_tokens: Approximate budget for returned content. Snippets are
                accumulated until this budget (≈4 chars/token) is exhausted.
            peer: Peer alias or explicit peer ID whose sessions to search.

        Returns:
            Ranked message excerpts as a formatted string, or empty string
            if none found.

        Raises:
            HonchoAuthError: rejected credentials, so no-results is not implied.
        """
        session = self._cache.get(session_key)
        if not session:
            return ""

        # peer_perspective spans the target peer's sessions across all authors.
        peer_id = self._resolve_peer_id(session, peer)

        # Honcho caps query length for the embedding model; keep well under it.
        q = (query or "").strip()
        if not q:
            return ""
        if len(q) > 4000:
            q = q[:4000]

        # Approximate four characters per token and a few hundred per result.
        char_budget = max(200, int(max_tokens) * 4)
        limit = max(3, min(20, char_budget // 300))

        try:
            messages = self._authed_call(
                "message search",
                lambda: self.honcho.search(
                    q,
                    filters={"peer_perspective": peer_id},
                    limit=limit,
                ),
            )
        except HonchoAuthError:
            raise
        except Exception as e:
            logger.debug("Honcho message search failed (peer_perspective=%s): %s", peer_id, e)
            # Fall back to peer-authored search if the perspective filter is
            # unsupported by the running Honcho version.
            try:
                messages = self._authed_call(
                    "peer search",
                    lambda: self._get_or_create_peer(peer_id).search(q, limit=limit),
                )
            except HonchoAuthError:
                raise
            except Exception as e2:
                logger.debug("Honcho peer search fallback also failed: %s", e2)
                return ""

        if not messages:
            return ""

        # Author labels distinguish user-stated facts from assistant-derived ones.
        assistant_id = session.assistant_peer_id
        lines: list[str] = []
        used = 0
        for m in messages:
            content = (getattr(m, "content", "") or "").strip()
            if not content:
                continue
            author = getattr(m, "peer_id", "") or "unknown"
            who = "assistant" if author == assistant_id else author
            sess = getattr(m, "session_id", "") or ""
            snippet = content[:1200]
            entry = f"[{who}{f' · {sess}' if sess else ''}] {snippet}"
            separator = "\n\n" if lines else ""
            remaining = char_budget - used - len(separator)
            if remaining <= 0:
                break
            if len(entry) > remaining:
                entry = entry[:remaining].rstrip()
                if not entry:
                    break
                lines.append(entry)
                used += len(separator) + len(entry)
                break
            lines.append(entry)
            used += len(separator) + len(entry)

        return "\n\n".join(lines)

    def _conclusions_scope(self, session: Any, target_peer_id: str) -> Any:
        """Resolve the ConclusionScope for observing target_peer_id.

        Shared by create/delete/list_conclusions so the observer/observed
        routing (self-conclusions vs. AI-observes-others vs. peer-owned)
        stays consistent across all three.
        """
        if target_peer_id == session.assistant_peer_id:
            observer = self._get_or_create_peer(session.assistant_peer_id)
            return observer.conclusions_of(session.assistant_peer_id)
        elif self._ai_observe_others:
            observer = self._get_or_create_peer(session.assistant_peer_id)
            return observer.conclusions_of(target_peer_id)
        else:
            target_peer = self._get_or_create_peer(target_peer_id)
            return target_peer.conclusions_of(target_peer_id)

    def create_conclusion(self, session_key: str, content: str, peer: str = "user") -> bool:
        """Write a conclusion about a target peer back to Honcho.

        Conclusions are facts a peer observes about another peer or itself —
        preferences, corrections, clarifications, and project context.
        They feed into the target peer's card and representation.

        Args:
            session_key: Session to associate the conclusion with.
            content: The conclusion text.
            peer: Peer alias or explicit peer ID. "user" is the default alias.

        Returns:
            True on success, False on failure.
        """
        if not content or not content.strip():
            return False

        session = self._cache.get(session_key)
        if not session:
            logger.warning("No session cached for '%s', skipping conclusion", session_key)
            return False

        try:
            target_peer_id = self._resolve_peer_id(session, peer)
            if target_peer_id is None:
                logger.warning("Could not resolve conclusion peer '%s' for session '%s'", peer, session_key)
                return False

            self._authed_call(
                "conclusion create",
                lambda: self._conclusions_scope(session, target_peer_id).create([{
                    "content": content.strip(),
                    "session_id": session.honcho_session_id,
                }]),
            )
            logger.info("Created conclusion about %s for %s: %s", target_peer_id, session_key, content[:80])
            return True
        except HonchoAuthError:
            raise
        except Exception as e:
            logger.error("Failed to create conclusion: %s", e)
            return False

    def delete_conclusion(self, session_key: str, conclusion_id: str, peer: str = "user") -> bool:
        """Delete a conclusion by ID. Use only for PII removal.

        Args:
            session_key: Session key for peer resolution.
            conclusion_id: The conclusion ID to delete.
            peer: Peer alias or explicit peer ID.

        Returns:
            True on success, False on failure.
        """
        session = self._cache.get(session_key)
        if not session:
            return False
        try:
            target_peer_id = self._resolve_peer_id(session, peer)
            self._authed_call(
                "conclusion delete",
                lambda: self._conclusions_scope(session, target_peer_id).delete(conclusion_id),
            )
            logger.info("Deleted conclusion %s for %s", conclusion_id, session_key)
            return True
        except HonchoAuthError:
            raise
        except Exception as e:
            logger.error("Failed to delete conclusion %s: %s", conclusion_id, e)
            return False

    def list_conclusions(
        self,
        session_key: str,
        query: str | None = None,
        peer: str = "user",
        limit: int = 20,
    ) -> list[dict]:
        """List or semantically search conclusions with their server IDs.

        Args:
            session_key: Session key for peer resolution.
            query: Optional semantic search query. Omit to list recent conclusions.
            peer: Peer alias or explicit peer ID.
            limit: Max conclusions to return.

        Returns:
            List of {"id": ..., "content": ...} dicts, or [] on failure/no session.
        """
        session = self._cache.get(session_key)
        if not session:
            return []
        try:
            target_peer_id = self._resolve_peer_id(session, peer)
            if target_peer_id is None:
                return []

            def _list() -> Any:
                scope = self._conclusions_scope(session, target_peer_id)
                if query:
                    return scope.query(query, top_k=limit)
                return scope.list(size=limit).items

            conclusions = self._authed_call("conclusion list", _list)
            return [{"id": c.id, "content": c.content} for c in conclusions]
        except HonchoAuthError:
            raise
        except Exception as e:
            logger.debug("Honcho list_conclusions failed: %s", e)
            return []

    def set_peer_card(self, session_key: str, card: list[str], peer: str = "user") -> list[str] | None:
        """Update a peer's card.

        Args:
            session_key: Session key for peer resolution.
            card: New peer card as list of fact strings.
            peer: Peer alias or explicit peer ID.

        Returns:
            Updated card on success, None on failure.
        """
        session = self._cache.get(session_key)
        if not session:
            return None
        try:
            observer_peer_id, target_peer_id = self._resolve_observer_target(session, peer)
            if observer_peer_id is None:
                logger.warning("Could not resolve peer '%s' for set_peer_card in session '%s'", peer, session_key)
                return None

            def _set_card() -> Any:
                peer_obj = self._get_or_create_peer(observer_peer_id)
                if target_peer_id is not None:
                    return peer_obj.set_card(card, target=target_peer_id)
                return peer_obj.set_card(card)

            result = self._authed_call("peer card update", _set_card)
            logger.info(
                "Updated peer card observer=%s target=%s (%d facts)",
                observer_peer_id,
                target_peer_id or observer_peer_id,
                len(card),
            )
            return result
        except HonchoAuthError:
            raise
        except Exception as e:
            logger.error("Failed to set peer card: %s", e)
            return None

    def seed_ai_identity(self, session_key: str, content: str, source: str = "manual") -> bool:
        """
        Seed the AI peer's Honcho representation from text content.

        Useful for priming AI identity from SOUL.md, exported chats, or
        any structured description. The content is sent as an assistant
        peer message so Honcho's reasoning model can incorporate it.

        Args:
            session_key: The session key to associate with.
            content: The identity/persona content to seed.
            source: Metadata tag for the source (e.g. "soul_md", "export").

        Returns:
            True on success, False on failure.
        """
        if not content or not content.strip():
            return False

        session = self._cache.get(session_key)
        if not session:
            logger.warning("No session cached for '%s', skipping AI seed", session_key)
            return False

        if session.honcho_session_id not in self._sessions_cache:
            logger.warning("No Honcho session cached for '%s', skipping AI seed", session_key)
            return False

        try:
            wrapped = (
                f"<ai_identity_seed>\n"
                f"<source>{source}</source>\n"
                f"\n"
                f"{content.strip()}\n"
                f"</ai_identity_seed>"
            )

            def _seed() -> None:
                assistant_peer = self._get_or_create_peer(session.assistant_peer_id)
                self._sdk_session(session.honcho_session_id).add_messages(
                    [assistant_peer.message(wrapped)]
                )

            self._authed_call("identity seed", _seed)
            logger.info("Seeded AI identity from '%s' into %s", source, session_key)
            return True
        except Exception as e:
            logger.error("Failed to seed AI identity: %s", e)
            return False

    def get_ai_representation(self, session_key: str) -> dict[str, str]:
        """
        Fetch the AI peer's current Honcho representation.

        Returns:
            Dict with 'representation' and 'card' keys, empty strings if unavailable.
        """
        session = self._cache.get(session_key)
        if not session:
            return {"representation": "", "card": ""}

        try:
            ctx = self._fetch_peer_context(session.assistant_peer_id, target=session.assistant_peer_id)
            return {
                "representation": ctx["representation"] or "",
                "card": "\n".join(ctx["card"]),
            }
        except HonchoAuthError:
            raise
        except Exception as e:
            logger.debug("Failed to fetch AI representation: %s", e)
            return {"representation": "", "card": ""}

    def list_sessions(self) -> list[dict[str, Any]]:
        """List all cached sessions."""
        return [
            {
                "key": s.key,
                "created_at": s.created_at.isoformat(),
                "updated_at": s.updated_at.isoformat(),
                "message_count": len(s.messages),
            }
            for s in self._cache.values()
        ]