glaip-sdk 0.0.0b99__py3-none-any.whl

glaip_sdk/client/agents.py

@@ -0,0 +1,1353 @@
+#!/usr/bin/env python3
+"""Agent client for AIP SDK.
+
+Authors:
+    Raymond Christopher (raymond.christopher@gdplabs.id)
+    Christian Trisno Sen Long Chen (christian.t.s.l.chen@gdplabs.id)
+"""
+
+import asyncio
+import json
+import logging
+from collections.abc import AsyncGenerator, Callable, Iterator, Mapping
+from contextlib import asynccontextmanager
+from os import PathLike
+from pathlib import Path
+from typing import TYPE_CHECKING, Any, BinaryIO
+
+if TYPE_CHECKING:
+    from glaip_sdk.client.schedules import ScheduleClient
+
+import httpx
+from glaip_sdk.agents import Agent
+from glaip_sdk.client.payloads.agent import (
+    AgentCreateRequest,
+    AgentListParams,
+    AgentListResult,
+    AgentUpdateRequest,
+)
+from glaip_sdk.client.agent_runs import AgentRunsClient
+from glaip_sdk.client.base import BaseClient
+from glaip_sdk.client.mcps import MCPClient
+from glaip_sdk.client.run_rendering import (
+    AgentRunRenderingManager,
+    compute_timeout_seconds,
+)
+from glaip_sdk.client.shared import build_shared_config
+from glaip_sdk.client.tools import ToolClient
+from glaip_sdk.config.constants import (
+    AGENT_CONFIG_FIELDS,
+    DEFAULT_AGENT_FRAMEWORK,
+    DEFAULT_AGENT_RUN_TIMEOUT,
+    DEFAULT_AGENT_TYPE,
+    DEFAULT_AGENT_VERSION,
+    DEFAULT_MODEL,
+)
+from glaip_sdk.exceptions import NotFoundError, ValidationError
+from glaip_sdk.models import AgentResponse
+from glaip_sdk.payload_schemas.agent import list_server_only_fields
+from glaip_sdk.utils.agent_config import normalize_agent_config_for_import
+from glaip_sdk.utils.client_utils import (
+    aiter_sse_events,
+    create_model_instances,
+    find_by_name,
+    prepare_multipart_data,
+)
+from glaip_sdk.utils.import_export import (
+    convert_export_to_import_format,
+    merge_import_with_cli_args,
+)
+from glaip_sdk.utils.rendering.renderer import RichStreamRenderer
+from glaip_sdk.utils.resource_refs import is_uuid
+from glaip_sdk.utils.serialization import load_resource_from_file
+from glaip_sdk.utils.validation import validate_agent_instruction
+
+# API endpoints
+AGENTS_ENDPOINT = "/agents/"
+
+# SSE content type
+SSE_CONTENT_TYPE = "text/event-stream"
+
+# Set up module-level logger
+logger = logging.getLogger("glaip_sdk.agents")
+
+_SERVER_ONLY_IMPORT_FIELDS = set(list_server_only_fields()) | {"success", "message"}
+_MERGED_SEQUENCE_FIELDS = ("tools", "agents", "mcps")
+_DEFAULT_METADATA_TYPE = "custom"
+
+
+@asynccontextmanager
+async def _async_timeout_guard(
+    timeout_seconds: float | None,
+) -> AsyncGenerator[None, None]:
+    """Apply an asyncio timeout when a custom timeout is provided."""
+    if timeout_seconds is None:
+        yield
+        return
+    try:
+        async with asyncio.timeout(timeout_seconds):
+            yield
+    except asyncio.TimeoutError as exc:
+        raise httpx.TimeoutException(f"Request timed out after {timeout_seconds}s") from exc
+
+
+def _normalise_sequence(value: Any) -> list[Any] | None:
+    """Normalise optional sequence inputs to plain lists."""
+    if value is None:
+        return None
+    if isinstance(value, list):
+        return value
+    if isinstance(value, (tuple, set)):
+        return list(value)
+    return [value]
+
+
+def _normalise_sequence_fields(mapping: dict[str, Any]) -> None:
+    """Normalise merged sequence fields in-place."""
+    for field in _MERGED_SEQUENCE_FIELDS:
+        if field in mapping:
+            normalised = _normalise_sequence(mapping[field])
+            if normalised is not None:
+                mapping[field] = normalised
+
+
+def _merge_override_maps(
+    base_values: Mapping[str, Any],
+    extra_values: Mapping[str, Any],
+) -> dict[str, Any]:
+    """Merge override mappings while normalising sequence fields."""
+    merged: dict[str, Any] = {}
+    for source in (base_values, extra_values):
+        for key, value in source.items():
+            if value is None:
+                continue
+            merged[key] = _normalise_sequence(value) if key in _MERGED_SEQUENCE_FIELDS else value
+    return merged
+
+
+def _split_known_and_extra(
+    payload: Mapping[str, Any],
+    known_fields: Mapping[str, Any],
+) -> tuple[dict[str, Any], dict[str, Any]]:
+    """Split payload mapping into known request fields and extras."""
+    known: dict[str, Any] = {}
+    extras: dict[str, Any] = {}
+    for key, value in payload.items():
+        if value is None:
+            continue
+        if key in known_fields:
+            known[key] = value
+        else:
+            extras[key] = value
+    return known, extras
+
+
+def _prepare_agent_metadata(value: Any) -> dict[str, Any]:
+    """Ensure agent metadata contains ``type: custom`` by default."""
+    if value is None:
+        return {"type": _DEFAULT_METADATA_TYPE}
+    if not isinstance(value, Mapping):
+        return {"type": _DEFAULT_METADATA_TYPE}
+
+    prepared = dict(value)
+    metadata_type = prepared.get("type")
+    if not metadata_type:
+        prepared["type"] = _DEFAULT_METADATA_TYPE
+    return prepared
+
+
+def _load_agent_file_payload(file_path: Path, *, model_override: str | None) -> dict[str, Any]:
+    """Load agent configuration from disk and normalise legacy fields."""
+    if not file_path.exists():
+        raise FileNotFoundError(f"Agent configuration file not found: {file_path}")
+    if not file_path.is_file():
+        raise ValueError(f"Agent configuration path must point to a file: {file_path}")
+
+    raw_data = load_resource_from_file(file_path)
+    if not isinstance(raw_data, Mapping):
+        raise ValueError("Agent configuration file must contain a mapping/object.")
+
+    payload = convert_export_to_import_format(dict(raw_data))
+    payload = normalize_agent_config_for_import(payload, model_override)
+
+    for field in _SERVER_ONLY_IMPORT_FIELDS:
+        payload.pop(field, None)
+
+    return payload
+
+
+def _prepare_import_payload(
+    file_path: Path,
+    overrides: Mapping[str, Any],
+    *,
+    drop_model_fields: bool = False,
+) -> dict[str, Any]:
+    """Prepare merged payload from file contents and explicit overrides."""
+    overrides_dict = dict(overrides)
+
+    raw_definition = load_resource_from_file(file_path)
+    original_refs = _extract_original_refs(raw_definition)
+
+    base_payload = _load_agent_file_payload(file_path, model_override=overrides_dict.get("model"))
+
+    cli_args = _build_cli_args(overrides_dict)
+
+    merged = merge_import_with_cli_args(base_payload, cli_args)
+
+    additional = _build_additional_args(overrides_dict, cli_args)
+    merged.update(additional)
+
+    if drop_model_fields:
+        _remove_model_fields_if_needed(merged, overrides_dict)
+
+    _set_default_refs(merged, original_refs)
+
+    _normalise_sequence_fields(merged)
+    return merged
+
+
+def _extract_original_refs(raw_definition: dict) -> dict[str, list]:
+    """Extract original tool/agent/mcp references from raw definition."""
+    return {
+        "tools": list(raw_definition.get("tools") or []),
+        "agents": list(raw_definition.get("agents") or []),
+        "mcps": list(raw_definition.get("mcps") or []),
+    }
+
+
+def _build_cli_args(overrides_dict: dict) -> dict[str, Any]:
+    """Build CLI args from overrides, filtering out None values."""
+    cli_args = {key: overrides_dict.get(key) for key in AGENT_CONFIG_FIELDS if overrides_dict.get(key) is not None}
+
+    # Normalize sequence fields
+    for field in _MERGED_SEQUENCE_FIELDS:
+        if field in cli_args:
+            cli_args[field] = tuple(_normalise_sequence(cli_args[field]) or [])
+
+    return cli_args
+
+
+def _build_additional_args(overrides_dict: dict, cli_args: dict) -> dict[str, Any]:
+    """Build additional args not already in CLI args."""
+    return {key: value for key, value in overrides_dict.items() if value is not None and key not in cli_args}
+
+
+def _remove_model_fields_if_needed(merged: dict, overrides_dict: dict) -> None:
+    """Remove model fields if not explicitly overridden."""
+    if overrides_dict.get("language_model_id") is None:
+        merged.pop("language_model_id", None)
+    if overrides_dict.get("provider") is None:
+        merged.pop("provider", None)
+
+
+def _set_default_refs(merged: dict, original_refs: dict) -> None:
+    """Set default references if not already present."""
+    merged.setdefault("_tool_refs", original_refs["tools"])
+    merged.setdefault("_agent_refs", original_refs["agents"])
+    merged.setdefault("_mcp_refs", original_refs["mcps"])
+
+
+class AgentClient(BaseClient):
+    """Client for agent operations."""
+
+    def __init__(
+        self,
+        *,
+        parent_client: BaseClient | None = None,
+        **kwargs: Any,
+    ) -> None:
+        """Initialize the agent client.
+
+        Args:
+            parent_client: Parent client to adopt session/config from
+            **kwargs: Additional arguments for standalone initialization
+        """
+        super().__init__(parent_client=parent_client, **kwargs)
+        self._renderer_manager = AgentRunRenderingManager(logger)
+        self._tool_client: ToolClient | None = None
+        self._mcp_client: MCPClient | None = None
+        self._runs_client: AgentRunsClient | None = None
+        self._schedule_client: ScheduleClient | None = None
+
+    def list_agents(
+        self,
+        query: AgentListParams | None = None,
+        **kwargs: Any,
+    ) -> AgentListResult:
+        """List agents with optional filtering and pagination support.
+
+        Args:
+            query: Query parameters for filtering agents. If None, uses kwargs to create query.
+            **kwargs: Individual filter parameters for backward compatibility.
+        """
+        if query is not None and kwargs:
+            # Both query object and individual parameters provided
+            raise ValueError("Provide either `query` or individual filter arguments, not both.")
+
+        if query is None:
+            # Create query from individual parameters for backward compatibility
+            query = AgentListParams(**kwargs)
+
+        params = query.to_query_params()
+        envelope = self._request_with_envelope(
+            "GET",
+            AGENTS_ENDPOINT,
+            params=params if params else None,
+        )
+
+        if not isinstance(envelope, dict):
+            envelope = {"data": envelope}
+
+        data_payload = envelope.get("data") or []
+        items = create_model_instances(data_payload, Agent, self)
+
+        return AgentListResult(
+            items=items,
+            total=envelope.get("total"),
+            page=envelope.get("page"),
+            limit=envelope.get("limit"),
+            has_next=envelope.get("has_next"),
+            has_prev=envelope.get("has_prev"),
+            message=envelope.get("message"),
+        )
+
+    def sync_langflow_agents(
+        self,
+        base_url: str | None = None,
+        api_key: str | None = None,
+    ) -> dict[str, Any]:
+        """Sync LangFlow agents by fetching flows from the LangFlow server.
+
+        This method synchronizes agents with LangFlow flows. It fetches all flows
+        from the configured LangFlow server and creates/updates corresponding agents.
+
+        Args:
+            base_url: Custom LangFlow server base URL. If not provided, uses LANGFLOW_BASE_URL env var.
+            api_key: Custom LangFlow API key. If not provided, uses LANGFLOW_API_KEY env var.
+
+        Returns:
+            Response containing sync results and statistics
+
+        Raises:
+            ValueError: If LangFlow server configuration is missing
+        """
+        payload = {}
+        if base_url is not None:
+            payload["base_url"] = base_url
+        if api_key is not None:
+            payload["api_key"] = api_key
+
+        return self._request("POST", "/agents/langflow/sync", json=payload)
+
+    def get_agent_by_id(self, agent_id: str) -> Agent:
+        """Get agent by ID."""
+        try:
+            data = self._request("GET", f"/agents/{agent_id}")
+        except ValidationError as exc:
+            if exc.status_code == 422:
+                message = f"Agent '{agent_id}' not found"
+                raise NotFoundError(
+                    message,
+                    status_code=404,
+                    error_type=exc.error_type,
+                    payload=exc.payload,
+                    request_id=exc.request_id,
+                ) from exc
+            raise
+
+        if isinstance(data, str):
+            # Some backends may respond with plain text for missing agents.
+            message = data.strip() or f"Agent '{agent_id}' not found"
+            raise NotFoundError(message, status_code=404)
+
+        if not isinstance(data, dict):
+            raise NotFoundError(
+                f"Agent '{agent_id}' not found (unexpected response type)",
+                status_code=404,
+            )
+
+        response = AgentResponse(**data)
+        return Agent.from_response(response, client=self)
+
+    def find_agents(self, name: str | None = None) -> list[Agent]:
+        """Find agents by name."""
+        result = self.list_agents(name=name)
+        agents = list(result)
+        if name is None:
+            return agents
+        return find_by_name(agents, name, case_sensitive=False)
+
+    # ------------------------------------------------------------------ #
+    # Renderer delegation helpers
+    # ------------------------------------------------------------------ #
+    def _get_renderer_manager(self) -> AgentRunRenderingManager:
+        """Get or create the renderer manager instance.
+
+        Returns:
+            AgentRunRenderingManager instance.
+        """
+        manager = getattr(self, "_renderer_manager", None)
+        if manager is None:
+            manager = AgentRunRenderingManager(logger)
+            self._renderer_manager = manager
+        return manager
+
+    def _create_renderer(self, renderer: RichStreamRenderer | str | None, **kwargs: Any) -> RichStreamRenderer:
+        """Create or return a renderer instance.
+
+        Args:
+            renderer: Renderer instance, string identifier, or None.
+            **kwargs: Additional keyword arguments (e.g., verbose).
+
+        Returns:
+            RichStreamRenderer instance.
+        """
+        manager = self._get_renderer_manager()
+        verbose = kwargs.get("verbose", False)
+        if isinstance(renderer, RichStreamRenderer) or hasattr(renderer, "on_start"):
+            return renderer  # type: ignore[return-value]
+        return manager.create_renderer(renderer, verbose=verbose)
+
+    def _process_stream_events(
+        self,
+        stream_response: httpx.Response,
+        renderer: RichStreamRenderer,
+        timeout_seconds: float,
+        agent_name: str | None,
+        meta: dict[str, Any],
+    ) -> tuple[str, dict[str, Any], float | None, float | None]:
+        """Process stream events from an HTTP response.
+
+        Args:
+            stream_response: HTTP response stream.
+            renderer: Renderer to use for displaying events.
+            timeout_seconds: Timeout in seconds.
+            agent_name: Optional agent name.
+            meta: Metadata dictionary.
+
+        Returns:
+            Tuple of (final_text, stats_usage, started_monotonic, finished_monotonic).
+        """
+        manager = self._get_renderer_manager()
+        return manager.process_stream_events(
+            stream_response,
+            renderer,
+            timeout_seconds,
+            agent_name,
+            meta,
+        )
+
+    def _finalize_renderer(
+        self,
+        renderer: RichStreamRenderer,
+        final_text: str,
+        stats_usage: dict[str, Any],
+        started_monotonic: float | None,
+        finished_monotonic: float | None,
+    ) -> str:
+        """Finalize the renderer and return the final response text.
+
+        Args:
+            renderer: Renderer to finalize.
+            final_text: Final text content.
+            stats_usage: Usage statistics dictionary.
+            started_monotonic: Start time (monotonic).
+            finished_monotonic: Finish time (monotonic).
+
+        Returns:
+            Final text string.
+        """
+        from glaip_sdk.client.run_rendering import finalize_render_manager  # noqa: PLC0415
+
+        manager = self._get_renderer_manager()
+        return finalize_render_manager(
+            manager, renderer, final_text, stats_usage, started_monotonic, finished_monotonic
+        )
+
+    def _get_tool_client(self) -> ToolClient:
+        """Get or create the tool client instance.
+
+        Returns:
+            ToolClient instance.
+        """
+        if self._tool_client is None:
+            self._tool_client = ToolClient(parent_client=self)
+        return self._tool_client
+
+    def _get_mcp_client(self) -> MCPClient:
+        """Get or create the MCP client instance.
+
+        Returns:
+            MCPClient instance.
+        """
+        if self._mcp_client is None:
+            self._mcp_client = MCPClient(parent_client=self)
+        return self._mcp_client
+
+    @property
+    def schedules(self) -> "ScheduleClient":
+        """Get or create the schedule client instance.
+
+        Returns:
+            ScheduleClient instance.
+        """
+        if self._schedule_client is None:
+            # Import here to avoid circular import
+            from glaip_sdk.client.schedules import ScheduleClient  # noqa: PLC0415
+
+            self._schedule_client = ScheduleClient(parent_client=self)
+        return self._schedule_client
+
+    def _normalise_reference_entry(
+        self,
+        entry: Any,
+        fallback_iter: Iterator[Any] | None,
+    ) -> tuple[str | None, str | None]:
+        """Normalize a reference entry to extract ID and name.
+
+        Args:
+            entry: Reference entry (string, dict, or other).
+            fallback_iter: Optional iterator for fallback values.
+
+        Returns:
+            Tuple of (entry_id, entry_name).
+        """
+        entry_id: str | None = None
+        entry_name: str | None = None
+
+        if isinstance(entry, str):
+            if is_uuid(entry):
+                entry_id = entry
+            else:
+                entry_name = entry
+        elif isinstance(entry, dict):
+            entry_id = entry.get("id")
+            entry_name = entry.get("name")
+        else:
+            entry_name = str(entry)
+
+        if entry_name or fallback_iter is None:
+            return entry_id, entry_name
+
+        try:
+            ref = next(fallback_iter)
+        except StopIteration:
+            ref = None
+        if isinstance(ref, dict):
+            entry_name = ref.get("name") or entry_name
+
+        return entry_id, entry_name
+
+    def _resolve_resource_ids(
+        self,
+        items: list[Any] | None,
+        references: list[Any] | None,
+        *,
+        fetch_by_id: Callable[[str], Any],
+        find_by_name: Callable[[str], list[Any]],
+        label: str,
+        plural_label: str | None = None,
+    ) -> list[str] | None:
+        """Resolve a list of resource references to IDs.
+
+        Args:
+            items: List of resource references to resolve.
+            references: Optional list of reference objects for fallback.
+            fetch_by_id: Function to fetch resource by ID.
+            find_by_name: Function to find resources by name.
+            label: Singular label for error messages.
+            plural_label: Plural label for error messages.
+
+        Returns:
+            List of resolved resource IDs, or None if items is empty.
+        """
+        if not items:
+            return None
+
+        if references is None:
+            return [self._coerce_reference_value(entry) for entry in items]
+
+        singular = label
+        plural = plural_label or f"{label}s"
+        fallback_iter = iter(references or [])
+
+        return [
+            self._resolve_single_resource(
+                entry,
+                fallback_iter,
+                fetch_by_id,
+                find_by_name,
+                singular,
+                plural,
+            )
+            for entry in items
+        ]
+
+    def _resolve_single_resource(
+        self,
+        entry: Any,
+        fallback_iter: Iterator[Any] | None,
+        fetch_by_id: Callable[[str], Any],
+        find_by_name: Callable[[str], list[Any]],
+        singular: str,
+        plural: str,
+    ) -> str:
+        """Resolve a single resource reference to an ID.
+
+        Args:
+            entry: Resource reference to resolve.
+            fallback_iter: Optional iterator for fallback values.
+            fetch_by_id: Function to fetch resource by ID.
+            find_by_name: Function to find resources by name.
+            singular: Singular label for error messages.
+            plural: Plural label for error messages.
+
+        Returns:
+            Resolved resource ID string.
+
+        Raises:
+            ValueError: If the resource cannot be resolved.
+        """
+        entry_id, entry_name = self._normalise_reference_entry(entry, fallback_iter)
+
+        validated_id = self._validate_resource_id(fetch_by_id, entry_id)
+        if validated_id:
+            return validated_id
+        if entry_id and entry_name is None:
+            return entry_id
+
+        if entry_name:
+            resolved, success = self._resolve_resource_by_name(find_by_name, entry_name, singular, plural)
+            return resolved if success else entry_name
+
+        raise ValueError(f"{singular} references must include a valid ID or name.")
+
+    @staticmethod
+    def _coerce_reference_value(entry: Any) -> str:
+        """Coerce a reference entry to a string value.
+
+        Args:
+            entry: Reference entry (dict, string, or other).
+
+        Returns:
+            String representation of the reference.
+        """
+        if isinstance(entry, dict):
+            if entry.get("id"):
+                return str(entry["id"])
+            if entry.get("name"):
+                return str(entry["name"])
+        return str(entry)
+
+    @staticmethod
+    def _validate_resource_id(fetch_by_id: Callable[[str], Any], candidate_id: str | None) -> str | None:
+        """Validate a resource ID by attempting to fetch it.
+
+        Args:
+            fetch_by_id: Function to fetch resource by ID.
+            candidate_id: Candidate ID to validate.
+
+        Returns:
+            Validated ID if found, None otherwise.
+        """
+        if not candidate_id:
+            return None
+        try:
+            fetch_by_id(candidate_id)
+        except Exception:
+            return None
+        return candidate_id
+
+    @staticmethod
+    def _resolve_resource_by_name(
+        find_by_name: Callable[[str], list[Any]],
+        entry_name: str,
+        singular: str,
+        plural: str,
+    ) -> tuple[str, bool]:
+        """Resolve a resource by name to an ID.
+
+        Args:
+            find_by_name: Function to find resources by name.
+            entry_name: Name of the resource to find.
+            singular: Singular label for error messages.
+            plural: Plural label for error messages.
+
+        Returns:
+            Tuple of (resolved_id, success).
+
+        Raises:
+            ValueError: If resource not found or multiple matches exist.
+        """
+        try:
+            matches = find_by_name(entry_name)
+        except Exception:
+            return entry_name, False
+
+        if not matches:
+            raise ValueError(f"{singular} '{entry_name}' not found in current workspace.")
+        if len(matches) > 1:
+            exact = [m for m in matches if getattr(m, "name", "").lower() == entry_name.lower()]
+            if len(exact) == 1:
+                matches = exact
+            else:
+                raise ValueError(f"Multiple {plural} named '{entry_name}'. Please disambiguate.")
+        return str(matches[0].id), True
+
+    def _resolve_tool_ids(
+        self,
+        tools: list[Any] | None,
+        references: list[Any] | None = None,
+    ) -> list[str] | None:
+        """Resolve tool references to IDs.
+
+        Args:
+            tools: List of tool references to resolve.
+            references: Optional list of reference objects for fallback.
+
+        Returns:
+            List of resolved tool IDs, or None if tools is empty.
+        """
+        tool_client = self._get_tool_client()
+        return self._resolve_resource_ids(
+            tools,
+            references,
+            fetch_by_id=tool_client.get_tool_by_id,
+            find_by_name=tool_client.find_tools,
+            label="Tool",
+            plural_label="tools",
+        )
+
+    def _resolve_agent_ids(
+        self,
+        agents: list[Any] | None,
+        references: list[Any] | None = None,
+    ) -> list[str] | None:
+        """Resolve agent references to IDs.
+
+        Args:
+            agents: List of agent references to resolve.
+            references: Optional list of reference objects for fallback.
+
+        Returns:
+            List of resolved agent IDs, or None if agents is empty.
+        """
+        return self._resolve_resource_ids(
+            agents,
+            references,
+            fetch_by_id=self.get_agent_by_id,
+            find_by_name=self.find_agents,
+            label="Agent",
+            plural_label="agents",
+        )
+
+    def _resolve_mcp_ids(
+        self,
+        mcps: list[Any] | None,
+        references: list[Any] | None = None,
+    ) -> list[str] | None:
+        """Resolve MCP references to IDs.
+
+        Args:
+            mcps: List of MCP references to resolve.
+            references: Optional list of reference objects for fallback.
+
+        Returns:
+            List of resolved MCP IDs, or None if mcps is empty.
+        """
+        mcp_client = self._get_mcp_client()
+        return self._resolve_resource_ids(
+            mcps,
+            references,
+            fetch_by_id=mcp_client.get_mcp_by_id,
+            find_by_name=mcp_client.find_mcps,
+            label="MCP",
+            plural_label="MCPs",
+        )
+
+    def _create_agent_from_payload(self, payload: Mapping[str, Any]) -> "Agent":
+        """Create an agent using a fully prepared payload mapping."""
+        known, extras = _split_known_and_extra(payload, AgentCreateRequest.__dataclass_fields__)
+
+        name = known.pop("name", None)
+        instruction = known.pop("instruction", None)
+        if not name or not str(name).strip():
+            raise ValueError("Agent name cannot be empty or whitespace")
+        if not instruction or not str(instruction).strip():
+            raise ValueError("Agent instruction cannot be empty or whitespace")
+
+        validated_instruction = validate_agent_instruction(str(instruction))
+        _normalise_sequence_fields(known)
+
+        resolved_model = known.pop("model", None) or DEFAULT_MODEL
+        tool_refs = extras.pop("_tool_refs", None)
+        agent_refs = extras.pop("_agent_refs", None)
+        mcp_refs = extras.pop("_mcp_refs", None)
+
+        tools_raw = known.pop("tools", None)
+        agents_raw = known.pop("agents", None)
+        mcps_raw = known.pop("mcps", None)
+
+        resolved_tools = self._resolve_tool_ids(tools_raw, tool_refs)
+        resolved_agents = self._resolve_agent_ids(agents_raw, agent_refs)
+        resolved_mcps = self._resolve_mcp_ids(mcps_raw, mcp_refs)
+
+        language_model_id = known.pop("language_model_id", None)
+        provider = known.pop("provider", None)
+        model_name = known.pop("model_name", None)
+
+        agent_type_value = known.pop("agent_type", None)
+        fallback_type_value = known.pop("type", None)
+        if agent_type_value is None:
+            agent_type_value = fallback_type_value or DEFAULT_AGENT_TYPE
+
+        framework_value = known.pop("framework", None) or DEFAULT_AGENT_FRAMEWORK
+        version_value = known.pop("version", None) or DEFAULT_AGENT_VERSION
+        account_id = known.pop("account_id", None)
+        description = known.pop("description", None)
+        metadata = _prepare_agent_metadata(known.pop("metadata", None))
+        tool_configs = known.pop("tool_configs", None)
+        agent_config = known.pop("agent_config", None)
+        timeout_value = known.pop("timeout", None)
+        a2a_profile = known.pop("a2a_profile", None)
+
+        final_extras = {**known, **extras}
+        final_extras.setdefault("model", resolved_model)
+
+        request = AgentCreateRequest(
+            name=str(name).strip(),
+            instruction=validated_instruction,
+            model=resolved_model,
+            language_model_id=language_model_id,
+            provider=provider,
+            model_name=model_name,
+            agent_type=agent_type_value,
+            framework=framework_value,
+            version=version_value,
+            account_id=account_id,
+            description=description,
+            metadata=metadata,
+            tools=resolved_tools,
+            agents=resolved_agents,
+            mcps=resolved_mcps,
+            tool_configs=tool_configs,
+            agent_config=agent_config,
+            timeout=timeout_value or DEFAULT_AGENT_RUN_TIMEOUT,
+            a2a_profile=a2a_profile,
+            extras=final_extras,
+        )
+
+        payload_dict = request.to_payload()
+        payload_dict.setdefault("model", resolved_model)
+
+        full_agent_data = self._post_then_fetch(
+            id_key="id",
+            post_endpoint=AGENTS_ENDPOINT,
+            get_endpoint_fmt=f"{AGENTS_ENDPOINT}{{id}}",
+            json=payload_dict,
+        )
+        response = AgentResponse(**full_agent_data)
+        return Agent.from_response(response, client=self)
+
+    def create_agent(
+        self,
+        name: str | None = None,
+        instruction: str | None = None,
+        model: str | None = None,
+        tools: list[str | Any] | None = None,
+        agents: list[str | Any] | None = None,
+        timeout: int | None = None,
+        *,
+        file: str | PathLike[str] | None = None,
+        mcps: list[str | Any] | None = None,
+        tool_configs: Mapping[str, Any] | None = None,
+        **kwargs: Any,
+    ) -> "Agent":
+        """Create a new agent, optionally loading configuration from a file."""
+        base_overrides = {
+            "name": name,
+            "instruction": instruction,
+            "model": model,
+            "tools": tools,
+            "agents": agents,
+            "timeout": timeout,
+            "mcps": mcps,
+            "tool_configs": tool_configs,
+        }
+        overrides = _merge_override_maps(base_overrides, kwargs)
+
+        if file is not None:
+            payload = _prepare_import_payload(Path(file).expanduser(), overrides, drop_model_fields=True)
+            if overrides.get("model") is None:
+                payload.pop("model", None)
+        else:
+            payload = overrides
+
+        return self._create_agent_from_payload(payload)
+
+    def create_agent_from_file(  # pragma: no cover - thin compatibility wrapper
+        self,
+        file_path: str | PathLike[str],
+        **overrides: Any,
+    ) -> "Agent":
+        """Backward-compatible helper to create an agent from a configuration file."""
+        return self.create_agent(file=file_path, **overrides)
+
+    def _update_agent_from_payload(
+        self,
+        agent_id: str,
+        current_agent: Agent,
+        payload: Mapping[str, Any],
+    ) -> "Agent":
+        """Update an agent using a prepared payload mapping."""
+        known, extras = _split_known_and_extra(payload, AgentUpdateRequest.__dataclass_fields__)
+        _normalise_sequence_fields(known)
+
+        tool_refs = extras.pop("_tool_refs", None)
+        agent_refs = extras.pop("_agent_refs", None)
+        mcp_refs = extras.pop("_mcp_refs", None)
+
+        tools_value = known.pop("tools", None)
+        agents_value = known.pop("agents", None)
+        mcps_value = known.pop("mcps", None)
+
+        if tools_value is not None:
+            tools_value = self._resolve_tool_ids(tools_value, tool_refs)
+        if agents_value is not None:
+            agents_value = self._resolve_agent_ids(agents_value, agent_refs)
+        if mcps_value is not None:
+            mcps_value = self._resolve_mcp_ids(mcps_value, mcp_refs)  # pragma: no cover
+
+        request = AgentUpdateRequest(
+            name=known.pop("name", None),
+            instruction=known.pop("instruction", None),
+            description=known.pop("description", None),
+            model=known.pop("model", None),
+            language_model_id=known.pop("language_model_id", None),
+            provider=known.pop("provider", None),
+            model_name=known.pop("model_name", None),
+            agent_type=known.pop("agent_type", known.pop("type", None)),
+            framework=known.pop("framework", None),
+            version=known.pop("version", None),
+            account_id=known.pop("account_id", None),
+            metadata=known.pop("metadata", None),
+            tools=tools_value,
+            tool_configs=known.pop("tool_configs", None),
+            agents=agents_value,
+            mcps=mcps_value,
+            agent_config=known.pop("agent_config", None),
+            a2a_profile=known.pop("a2a_profile", None),
+            extras={**known, **extras},
+        )
+
+        payload_dict = request.to_payload(current_agent)
+
+        api_response = self._request("PUT", f"/agents/{agent_id}", json=payload_dict)
+        response = AgentResponse(**api_response)
+        return Agent.from_response(response, client=self)
+
+    def update_agent(
+        self,
+        agent_id: str,
+        name: str | None = None,
+        instruction: str | None = None,
+        model: str | None = None,
+        *,
+        file: str | PathLike[str] | None = None,
+        tools: list[str | Any] | None = None,
+        agents: list[str | Any] | None = None,
+        mcps: list[str | Any] | None = None,
+        **kwargs: Any,
+    ) -> "Agent":
+        """Update an existing agent."""
+        base_overrides = {
+            "name": name,
+            "instruction": instruction,
+            "model": model,
+            "tools": tools,
+            "agents": agents,
+            "mcps": mcps,
+        }
+        overrides = _merge_override_maps(base_overrides, kwargs)
+
+        if file is not None:
+            payload = _prepare_import_payload(Path(file).expanduser(), overrides, drop_model_fields=True)
+        else:
+            payload = overrides
+
+        current_agent = self.get_agent_by_id(agent_id)
+        return self._update_agent_from_payload(agent_id, current_agent, payload)
+
+    def update_agent_from_file(  # pragma: no cover - thin compatibility wrapper
+        self,
+        agent_id: str,
+        file_path: str | PathLike[str],
+        **overrides: Any,
+    ) -> "Agent":
+        """Backward-compatible helper to update an agent from a configuration file."""
+        return self.update_agent(agent_id, file=file_path, **overrides)
+
+    def delete_agent(self, agent_id: str) -> None:
+        """Delete an agent."""
+        self._request("DELETE", f"/agents/{agent_id}")
+
+    def upsert_agent(self, identifier: str | Agent, **kwargs) -> Agent:
+        """Create or update an agent by instance, ID, or name.
+
+        Args:
+            identifier: Agent instance, ID (UUID string), or name
+            **kwargs: Agent configuration (instruction, description, tools, etc.)
+
+        Returns:
+            The created or updated agent.
+
+        Example:
+            >>> # By name (creates if not exists)
+            >>> agent = client.agents.upsert_agent(
+            ...     "hello_agent",
+            ...     instruction="You are a helpful assistant.",
+            ...     description="A friendly agent",
+            ... )
+            >>> # By instance
+            >>> agent = client.agents.upsert_agent(existing_agent, description="Updated")
+            >>> # By ID
+            >>> agent = client.agents.upsert_agent("uuid-here", description="Updated")
+        """
+        # Handle Agent instance
+        if isinstance(identifier, Agent):
+            if identifier.id:
+                logger.info("Updating agent by instance: %s", identifier.name)
+                return self.update_agent(identifier.id, name=identifier.name, **kwargs)
+            identifier = identifier.name
+
+        # Handle string (ID or name)
+        if isinstance(identifier, str):
+            # Check if it's a UUID
+            if is_uuid(identifier):
+                logger.info("Updating agent by ID: %s", identifier)
+                return self.update_agent(identifier, **kwargs)
+
+            # It's a name - find or create
+            return self._upsert_agent_by_name(identifier, **kwargs)
+
+        raise ValueError(f"Invalid identifier type: {type(identifier)}")
+
+    def _upsert_agent_by_name(self, name: str, **kwargs) -> Agent:
+        """Find agent by name and update, or create if not found."""
+        existing = self.find_agents(name)
+
+        if len(existing) == 1:
+            logger.info("Updating existing agent: %s", name)
+            return self.update_agent(existing[0].id, name=name, **kwargs)
+
+        if len(existing) > 1:
+            raise ValueError(f"Multiple agents found with name '{name}'")
+
+        # Create new agent
+        logger.info("Creating new agent: %s", name)
+        return self.create_agent(name=name, **kwargs)
+
+    def _prepare_sync_request_data(
+        self,
+        message: str,
+        files: list[str | BinaryIO] | None,
+        tty: bool,
+        **kwargs: Any,
+    ) -> tuple[dict | None, dict | None, list | None, dict, Any | None]:
+        """Prepare request data for synchronous agent runs with renderer support.
+
+        Args:
+            message: Message to send
+            files: Optional files to include
+            tty: Whether to enable TTY mode
+            **kwargs: Additional request parameters
+
+        Returns:
+            Tuple of (payload, data_payload, files_payload, headers, multipart_data)
+        """
+        headers = {"Accept": SSE_CONTENT_TYPE}
+
+        if files:
+            # Handle multipart data for file uploads
+            multipart_data = prepare_multipart_data(message, files)
+            if "chat_history" in kwargs and kwargs["chat_history"] is not None:
+                multipart_data.data["chat_history"] = kwargs["chat_history"]
+            if "pii_mapping" in kwargs and kwargs["pii_mapping"] is not None:
+                multipart_data.data["pii_mapping"] = kwargs["pii_mapping"]
+
+            return (
+                None,
+                multipart_data.data,
+                multipart_data.files,
+                headers,
+                multipart_data,
+            )
+        else:
+            # Simple JSON payload for text-only requests
+            payload = {"input": message, "stream": True, **kwargs}
+            if tty:
+                payload["tty"] = True
+            return payload, None, None, headers, None
+
+    def _get_timeout_values(self, timeout: float | None, **kwargs: Any) -> tuple[float, float]:
+        """Get request timeout and execution timeout values.
+
+        Args:
+            timeout: Request timeout (overrides instance timeout)
+            **kwargs: Additional parameters including execution timeout
+
+        Returns:
+            Tuple of (request_timeout, execution_timeout)
+        """
+        request_timeout = timeout or self.timeout
+        execution_timeout = kwargs.get("timeout", DEFAULT_AGENT_RUN_TIMEOUT)
+        return request_timeout, execution_timeout
+
+    def run_agent(
+        self,
+        agent_id: str,
+        message: str,
+        files: list[str | BinaryIO] | None = None,
+        tty: bool = False,
+        *,
+        renderer: RichStreamRenderer | str | None = "auto",
+        runtime_config: dict[str, Any] | None = None,
+        **kwargs,
+    ) -> str:
+        """Run an agent with a message, streaming via a renderer.
+
+        Args:
+            agent_id: The ID of the agent to run.
+            message: The message to send to the agent.
+            files: Optional list of files to include with the request.
+            tty: Whether to enable TTY mode.
+            renderer: Renderer for streaming output.
+            runtime_config: Optional runtime configuration for tools, MCPs, and agents.
+                Keys should be platform IDs. Example:
+                {
+                    "tool_configs": {"tool-id": {"param": "value"}},
+                    "mcp_configs": {"mcp-id": {"setting": "on"}},
+                    "agent_config": {"planning": True},
+                }
+            **kwargs: Additional arguments to pass to the run API.
+
+        Returns:
+            The agent's response as a string.
+        """
+        # Include runtime_config in kwargs only when caller hasn't already provided it
+        if runtime_config is not None and "runtime_config" not in kwargs:
+            kwargs["runtime_config"] = runtime_config
+        (
+            payload,
+            data_payload,
+            files_payload,
+            headers,
+            multipart_data,
+        ) = self._prepare_sync_request_data(message, files, tty, **kwargs)
+
+        render_manager = self._get_renderer_manager()
+        verbose = kwargs.get("verbose", False)
+        r = self._create_renderer(renderer, verbose=verbose)
+        meta = render_manager.build_initial_metadata(agent_id, message, kwargs)
+        render_manager.start_renderer(r, meta)
+
+        final_text = ""
+        stats_usage: dict[str, Any] = {}
+        started_monotonic: float | None = None
+        finished_monotonic: float | None = None
+
+        timeout_seconds = compute_timeout_seconds(kwargs)
+
+        try:
+            response = self.http_client.stream(
+                "POST",
+                f"/agents/{agent_id}/run",
+                json=payload,
+                data=data_payload,
+                files=files_payload,
+                headers=headers,
+                timeout=timeout_seconds,
+            )
+
+            with response as stream_response:
+                stream_response.raise_for_status()
+
+                agent_name = kwargs.get("agent_name")
+
+                (
+                    final_text,
+                    stats_usage,
+                    started_monotonic,
+                    finished_monotonic,
+                ) = self._process_stream_events(
+                    stream_response,
+                    r,
+                    timeout_seconds,
+                    agent_name,
+                    meta,
+                )
+
+        except KeyboardInterrupt:
+            try:
+                r.close()
+            finally:
+                raise
+        except Exception:
+            try:
+                r.close()
+            finally:
+                raise
+        finally:
+            if multipart_data:
+                multipart_data.close()
+
+        return self._finalize_renderer(
+            r,
+            final_text,
+            stats_usage,
+            started_monotonic,
+            finished_monotonic,
+        )
+
+    def _prepare_request_data(
+        self,
+        message: str,
+        files: list[str | BinaryIO] | None,
+        **kwargs,
+    ) -> tuple[dict | None, dict | None, dict | None, dict | None]:
+        """Prepare request data for async agent runs.
+
+        Returns:
+            Tuple of (payload, data_payload, files_payload, headers)
+        """
+        if files:
+            # Handle multipart data for file uploads
+            multipart_data = prepare_multipart_data(message, files)
+            # Inject optional multipart extras expected by backend
+            if "chat_history" in kwargs and kwargs["chat_history"] is not None:
+                multipart_data.data["chat_history"] = kwargs["chat_history"]
+            if "pii_mapping" in kwargs and kwargs["pii_mapping"] is not None:
+                multipart_data.data["pii_mapping"] = kwargs["pii_mapping"]
+
+            headers = {"Accept": SSE_CONTENT_TYPE}
+            return None, multipart_data.data, multipart_data.files, headers
+        else:
+            # Simple JSON payload for text-only requests
+            payload = {"input": message, "stream": True, **kwargs}
+            headers = {"Accept": SSE_CONTENT_TYPE}
+            return payload, None, None, headers
+
+    def _create_async_client_config(self, timeout: float | None, headers: dict | None) -> dict:
+        """Create async client configuration with proper headers and timeout."""
+        config = self._build_async_client(timeout or self.timeout)
+        if headers:
+            config["headers"] = {**config["headers"], **headers}
+        return config
+
+    async def _stream_agent_response(
+        self,
+        async_client: httpx.AsyncClient,
+        agent_id: str,
+        payload: dict | None,
+        data_payload: dict | None,
+        files_payload: dict | None,
+        headers: dict | None,
+        timeout_seconds: float,
+        agent_name: str | None,
+    ) -> AsyncGenerator[dict, None]:
+        """Stream the agent response and yield parsed JSON chunks."""
+        async with async_client.stream(
+            "POST",
+            f"/agents/{agent_id}/run",
+            json=payload,
+            data=data_payload,
+            files=files_payload,
+            headers=headers,
+        ) as stream_response:
+            stream_response.raise_for_status()
+
+            async for event in aiter_sse_events(stream_response, timeout_seconds, agent_name):
+                try:
+                    chunk = json.loads(event["data"])
+                    yield chunk
+                except json.JSONDecodeError:
+                    logger.debug("Non-JSON SSE fragment skipped")
+                    continue
+
+    async def arun_agent(
+        self,
+        agent_id: str,
+        message: str,
+        files: list[str | BinaryIO] | None = None,
+        *,
+        request_timeout: float | None = None,
+        runtime_config: dict[str, Any] | None = None,
+        **kwargs,
+    ) -> AsyncGenerator[dict, None]:
+        """Async run an agent with a message, yielding streaming JSON chunks.
+
+        Args:
+            agent_id: ID of the agent to run
+            message: Message to send to the agent
+            files: Optional list of files to include
+            request_timeout: Optional request timeout in seconds (defaults to client timeout)
+            runtime_config: Optional runtime configuration for tools, MCPs, and agents.
+                Keys should be platform IDs. Example:
+                {
+                    "tool_configs": {"tool-id": {"param": "value"}},
+                    "mcp_configs": {"mcp-id": {"setting": "on"}},
+                    "agent_config": {"planning": True},
+                }
+            **kwargs: Additional arguments (chat_history, pii_mapping, etc.)
+
+        Yields:
+            Dictionary containing parsed JSON chunks from the streaming response
+
+        Raises:
+            AgentTimeoutError: When agent execution times out
+            httpx.TimeoutException: When general timeout occurs
+            Exception: For other unexpected errors
+        """
+        # Include runtime_config in kwargs only when caller hasn't already provided it
+        if runtime_config is not None and "runtime_config" not in kwargs:
+            kwargs["runtime_config"] = runtime_config
+        # Derive timeout values for request/control flow
+        legacy_timeout = kwargs.get("timeout")
+        http_timeout_override = request_timeout if request_timeout is not None else legacy_timeout
+        http_timeout = http_timeout_override or self.timeout
+
+        # Prepare request data
+        payload, data_payload, files_payload, headers = self._prepare_request_data(message, files, **kwargs)
+
+        # Create async client configuration
+        async_client_config = self._create_async_client_config(http_timeout_override, headers)
+
+        # Get execution timeout for streaming control
+        timeout_seconds = kwargs.get("timeout", DEFAULT_AGENT_RUN_TIMEOUT)
+        agent_name = kwargs.get("agent_name")
+
+        async def _chunk_stream() -> AsyncGenerator[dict, None]:
+            async with httpx.AsyncClient(**async_client_config) as async_client:
+                async for chunk in self._stream_agent_response(
+                    async_client,
+                    agent_id,
+                    payload,
+                    data_payload,
+                    files_payload,
+                    headers,
+                    timeout_seconds,
+                    agent_name,
+                ):
+                    yield chunk
+
+        async with _async_timeout_guard(http_timeout):
+            async for chunk in _chunk_stream():
+                yield chunk
+
+    @property
+    def runs(self) -> "AgentRunsClient":
+        """Get the agent runs client."""
+        if self._runs_client is None:
+            shared_config = build_shared_config(self)
+            self._runs_client = AgentRunsClient(**shared_config)
+        return self._runs_client