codeer-cli 0.1.10__tar.gz → 0.1.12__tar.gz

{codeer_cli-0.1.10 → codeer_cli-0.1.12}/API_REFERENCE.md

@@ -3,8 +3,9 @@
 > This file is the request-shape reference for maintaining `codeer-cli`.
 
 The 9 stages below mirror the user-docs lifecycle (`agent-creation` →
-`optimization-loop` → `publish`). Every path is under `/api/v1/`. All endpoints
-authenticate via `x-api-key` from `CODEER_API_KEY`.
+`optimization-loop` → `publish`). Agent, KB, eval, history, and model paths are
+under `/api/v1/`; Chat create/message/read paths use `/api/v2/chats`. All
+endpoints authenticate via `x-api-key` from `CODEER_API_KEY`.
 
 Envelope: successful responses look like
 `{"error_code": 0, "message": "", "pagination": null, "data": <payload>}`.
@@ -26,6 +27,9 @@ need a default agent.
 - `/histories` uses **`limit` + `offset`** (NOT `page` / `page_size`).
   Default in `histories.list()` is `limit=500`. Backend hard-cap may be
   lower — check the response length.
+- `/api/v2/chats/{id}/messages` also uses `limit` + `offset`.
+  `chats.list_messages()` follows pages until exhaustion; its `limit` argument
+  is a page size, not a total-result cap.
 - `/agents/{id}/histories`, `/eval/agents/{id}/cases`, `/eval/evaluators` all
   return everything in one shot today (no pagination).
 - `order_by` defaults to `"desc"` (most recent first) on endpoints that
@@ -140,19 +144,21 @@ codeer kb files --kb-id <kb-id>
 codeer kb faq-create --context-object-id <snapshot-object-id> --question "..." --range 12:0-12:42 --dry-run
 ```
 
-## Stage 3 — Live Test on a specific version
+## Stage 3 — Live Test on a specific version (legacy V1 compatibility)
 
 | Method & path | Purpose |
 | --- | --- |
-| `POST /chats` | Create a new chat session bound to an agent |
-| `POST /chats/{chat_id}/messages` | Send a message; **SSE stream** of tool calls + reasoning + final text |
-| `GET /chats/{chat_id}/messages` | Read historical messages for a chat |
-| `POST /chats/{chat_id}/regenerate` | Re-run the last turn |
-| `POST /chats/{chat_id}/messages/{msg_id}/feedbacks` | Thumbs up/down on a reply |
-
-`POST /chats/.../messages` requires `agent_history_id` — this is the key hook
-for the apply → test → publish workflow. Pass the draft `AgentHistory.id` from
-`PUT /agents/{id}` to test an unpublished version.
+| `POST /api/v1/chats` | Create a legacy chat session bound to an agent |
+| `POST /api/v1/chats/{chat_id}/messages` | Send a version-pinned message through legacy SSE |
+| `GET /api/v1/chats/{chat_id}/messages` | Read legacy messages for a chat |
+| `POST /api/v1/chats/{chat_id}/regenerate` | Re-run the last turn |
+| `POST /api/v1/chats/{chat_id}/messages/{msg_id}/feedbacks` | Thumbs up/down on a reply |
+
+Legacy V1 `POST /chats/.../messages` accepts `agent_history_id` — this is the
+compatibility hook for the apply → test → publish workflow. Pass the draft
+`AgentHistory.id` from `PUT /agents/{id}` to test an unpublished version.
+The API-key external flow in Chat V2 accepts `agent_id` and runs its published
+version; it does not currently replace this draft-pinning path.
 
 ## Stage 4 — Version management
 
@@ -267,18 +273,21 @@ the public CLI.
 
 | Method & path | Purpose |
 | --- | --- |
-| `POST /chats` | Create a persisted history using an agent's current published version |
-| `POST /chats/{id}/messages` | Append a turn to an existing persisted history using the current published version |
+| `POST /api/v2/chats` | Create a persisted history using an agent's current published version |
+| `POST /api/v2/chats/{id}/messages` | Append a turn through structured SSE using the current published version |
+| `GET /api/v2/chats/{id}/messages` | Read persisted Chat V2 conversation parts |
 | `GET /histories?agent_id=X&feedback_filter=improve_feedback&external_user_id=…` | List conversations with filters |
 | `GET /histories/{id}` | Read one history's metadata |
-| `GET /histories/{id}/conversations` | Full conversation turns incl. tool calls |
+| `GET /histories/{id}/conversations` | Legacy compact conversation rows; not complete tool I/O |
 | `POST /histories/{hid}/conversations/{cid}/feedbacks` | Leave freeform improvement feedback |
 | `POST /histories/{hid}/conversations/{cid}/score` | Numeric score |
 
 The CLI exposes the first two operations as `codeer history create` and
-`codeer history send`. Non-streaming message requests default to a 120-second
-per-message timeout. A timeout has an uncertain write outcome, so read the
-history before retrying to avoid duplicate turns.
+`codeer history send`. Messages explicitly set `stream: true`, consume Chat V2
+SSE, and require `response.completed` before reporting success. Their
+per-message SSE read timeout defaults to 240 seconds. A timeout,
+`response.failed`, or disconnect before completion has an uncertain write
+outcome, so read the history before retrying to avoid duplicate turns.
 
 `feedback_filter` accepts the `FeedbackFilterType` enum values:
 `no_feedback`, `with_feedback`, `helpful_feedback`, `improve_feedback`.
@@ -357,7 +366,7 @@ said (scope, factuality, tool-use rules).
 
 ### 4. Agent version pinning works everywhere — use it
 
-Both `POST /chats/{id}/messages` (`agent_history_id` required) and
+Legacy V1 `POST /chats/{id}/messages` (`agent_history_id` required) and
 `POST /eval/trigger` (`agent_history_id` optional, null = live state) accept
 the draft history id. The apply-→-test-→-publish loop:
 
@@ -461,11 +470,12 @@ For bulk creation, `codeer eval cases-apply --attachments-dir <dir>` reads
 each case's `attachment_files: ["x.jpg"]` array, uploads, and attaches in one
 pass. Workspace scope is inferred from the API-key virtual user profile.
 
-### 11. Tool args + outputs are NOT persisted in history reads
+### 11. Tool args + outputs are not available in legacy V1 history reads
 
-Conversations have only three roles (`OpenAIChatRole = system | user | assistant`)
-— there is no `tool` role row. When you read a history, here's what you can
-and can't recover from each assistant turn:
+Legacy V1 `History` conversation rows have only three roles
+(`OpenAIChatRole = system | user | assistant`) — there is no `tool` role row.
+When you read those rows through `/histories/{id}/conversations`, here's what
+you can and can't recover from each assistant turn:
 
 | Recoverable | Where |
 | --- | --- |
@@ -481,10 +491,12 @@ and can't recover from each assistant turn:
 | Tool **outputs** (raw JSON returned by the tool) | same — stored only as derived `primary_sources` for retrieval tools |
 | Reasoning steps mid-turn | `meta.reasoning_steps` is currently always `null` |
 
-If you need full tool I/O, capture it at execution time via the chat SSE
-stream (`POST /chats/{id}/messages`), not from history reads. For after-the-
-fact analysis, the persisted shape is sufficient to surface tool-selection
-patterns, token costs, and which sources the agent ended up citing.
+Chat V2 improves this contract: structured SSE emits tool calls and returns as
+`response.part.created` / `response.part.completed`, and
+`GET /api/v2/chats/{id}/messages` reads the persisted conversation parts.
+Capture the SSE artifact with `--out` when exact event order matters; use the
+paginated V2 message read for persisted after-the-fact tool I/O. The legacy V1
+history read remains useful for compact turn-level compatibility only.
 
 ### 10. A KB has exactly ONE level of folders — no nesting
 

{codeer_cli-0.1.10 → codeer_cli-0.1.12}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codeer-cli
-Version: 0.1.10
+Version: 0.1.12
 Summary: Command line tools for managing Codeer agents over the Codeer API.
 Project-URL: Homepage, https://www.codeer.ai
 Author: Codeer.AI
@@ -172,15 +172,18 @@ Use this pattern during agent lifecycle work:
 ```bash
 codeer agent list
 codeer history list --agent <agent-id> --limit 50
-codeer history create --agent <agent-id> --message "Review this plan" --timeout 120
-codeer history send <history-id> --message "Use the recommended options" --timeout 120
+codeer history conversations <history-id> --out .codeer/current/history-<history-id>.json
+codeer history create --agent <agent-id> --message "Review this plan" --timeout 240
+codeer history send <history-id> --message "Use the recommended options" --timeout 240
 codeer eval run --agent <agent-id> --cases <case-ids> --evaluator <evaluator-id> --out .codeer/eval_run.json
 ```
 
 `history create` and `history send` use the agent's current published version.
-Their per-message timeout defaults to 120 seconds. If a write request times
-out, inspect the history before retrying: the server may have completed the
-turn after the client stopped waiting.
+They use Chat V2 structured SSE with `stream: true`; their per-message read
+timeout defaults to 240 seconds. Success requires a `response.completed`
+event. If the stream times out, reports `response.failed`, or disconnects
+early, inspect the history before retrying: the server may already have
+persisted the turn.
 
 Eval case label commands always operate on the active API-key workspace. They
 do not accept a workspace override; switch CLI profiles to target another
@@ -194,6 +197,11 @@ Flags:
   for raw eval results, full conversation turns, full rubric matrices, and
   other data that can grow with cases, versions, or turns.
 
+`history conversations` reads Chat V2 parts and follows all pages
+automatically. Its stdout is still a bounded summary; the `--out` artifact is
+the complete client-visible history, including tool calls/results,
+attachments, interactions, feedback, and passthrough metadata.
+
 Avoid piping large raw JSON directly into agent chat. Prefer `--out`, then ask
 the coding agent to inspect targeted summaries, IDs, failing cases, or selected
 snippets from the saved file.

{codeer_cli-0.1.10 → codeer_cli-0.1.12}/README.md

@@ -154,15 +154,18 @@ Use this pattern during agent lifecycle work:
 ```bash
 codeer agent list
 codeer history list --agent <agent-id> --limit 50
-codeer history create --agent <agent-id> --message "Review this plan" --timeout 120
-codeer history send <history-id> --message "Use the recommended options" --timeout 120
+codeer history conversations <history-id> --out .codeer/current/history-<history-id>.json
+codeer history create --agent <agent-id> --message "Review this plan" --timeout 240
+codeer history send <history-id> --message "Use the recommended options" --timeout 240
 codeer eval run --agent <agent-id> --cases <case-ids> --evaluator <evaluator-id> --out .codeer/eval_run.json
 ```
 
 `history create` and `history send` use the agent's current published version.
-Their per-message timeout defaults to 120 seconds. If a write request times
-out, inspect the history before retrying: the server may have completed the
-turn after the client stopped waiting.
+They use Chat V2 structured SSE with `stream: true`; their per-message read
+timeout defaults to 240 seconds. Success requires a `response.completed`
+event. If the stream times out, reports `response.failed`, or disconnects
+early, inspect the history before retrying: the server may already have
+persisted the turn.
 
 Eval case label commands always operate on the active API-key workspace. They
 do not accept a workspace override; switch CLI profiles to target another
@@ -176,6 +179,11 @@ Flags:
   for raw eval results, full conversation turns, full rubric matrices, and
   other data that can grow with cases, versions, or turns.
 
+`history conversations` reads Chat V2 parts and follows all pages
+automatically. Its stdout is still a bounded summary; the `--out` artifact is
+the complete client-visible history, including tool calls/results,
+attachments, interactions, feedback, and passthrough metadata.
+
 Avoid piping large raw JSON directly into agent chat. Prefer `--out`, then ask
 the coding agent to inspect targeted summaries, IDs, failing cases, or selected
 snippets from the saved file.

{codeer_cli-0.1.10 → codeer_cli-0.1.12}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "codeer-cli"
-version = "0.1.10"
+version = "0.1.12"
 description = "Command line tools for managing Codeer agents over the Codeer API."
 readme = "README.md"
 requires-python = ">=3.11"

codeer_cli-0.1.12/src/codeer_cli/chats.py

@@ -0,0 +1,237 @@
+"""Chat V2 creation, structured SSE responses, and persisted message reads."""
+
+from __future__ import annotations
+
+from typing import Any, Iterator, List, Optional
+
+from .client import CodeerClient, TransportError
+
+
+def create(
+    client: CodeerClient,
+    *,
+    agent_id: str,
+    title: Optional[str] = None,
+    external_user_id: Optional[str] = None,
+) -> dict:
+    body: dict[str, Any] = {
+        "agent_id": agent_id,
+        "name": title or "CLI conversation",
+    }
+    if external_user_id is not None:
+        body["external_user_id"] = external_user_id
+    return client.post("/chats", api_version="v2", json=body)
+
+
+def send_published_agent_message(
+    client: CodeerClient,
+    *,
+    chat_id: int,
+    message: str,
+    agent_id: str,
+    external_user_id: Optional[str] = None,
+    attachment_ids: Optional[List[str]] = None,
+    stream: bool = True,
+    timeout: Optional[float] = None,
+) -> Iterator[dict] | dict:
+    """Send a user message through the API-key external chat flow.
+
+    API-key chat endpoints use the agent's published version. They accept
+    ``agent_id`` rather than ``agent_history_id``.
+    """
+    body: dict[str, Any] = {"message": message, "agent_id": agent_id, "stream": stream}
+    if external_user_id is not None:
+        body["external_user_id"] = external_user_id
+    if attachment_ids:
+        body["attached_file_uuids"] = attachment_ids
+
+    path = f"/chats/{chat_id}/messages"
+    if stream:
+        return client.stream_sse(
+            "POST",
+            path,
+            api_version="v2",
+            json=body,
+            timeout=timeout,
+        )
+    return client.post(path, api_version="v2", json=body, timeout=timeout)
+
+
+def send_message(
+    client: CodeerClient,
+    *,
+    chat_id: int,
+    message: str,
+    agent_history_id: str,
+    attachment_ids: Optional[List[str]] = None,
+    stream: bool = True,
+    timeout: Optional[float] = None,
+) -> Iterator[dict] | dict:
+    """Use legacy Chat V1 to pin an unpublished agent version.
+
+    Chat V2's API-key external flow only accepts the published ``agent_id``.
+    Keep this low-level compatibility helper on V1 until V2 supports external
+    ``agent_history_id`` pinning. It remains streaming by default.
+    """
+    body: dict[str, Any] = {
+        "message": message,
+        "agent_history_id": agent_history_id,
+        "stream": stream,
+    }
+    if attachment_ids:
+        body["attachment_ids"] = attachment_ids
+
+    path = f"/chats/{chat_id}/messages"
+    if stream:
+        return client.stream_sse(
+            "POST",
+            path,
+            json=body,
+            timeout=timeout,
+        )
+    return client.post(path, json=body, timeout=timeout)
+
+
+def collect_stream(events: Iterator[dict]) -> dict:
+    """Collect a Chat V2 SSE stream and require an explicit completion event."""
+    raw_events: list[dict] = []
+    parts: list[dict] = []
+    interactions: list[dict] = []
+    text_deltas: list[str] = []
+    final_text: str | None = None
+    response_id: str | None = None
+    conversation_group_id: str | None = None
+    updated_title: str | None = None
+    completed = False
+
+    for event in events:
+        data = event.get("data")
+        if data == "[DONE]":
+            continue
+        raw_events.append(event)
+        if not isinstance(data, dict):
+            continue
+
+        event_type = str(data.get("type") or event.get("event") or "message")
+        response_id = str(data.get("response_id") or response_id or "") or None
+        conversation_group_id = str(
+            data.get("conversation_group_id") or conversation_group_id or ""
+        ) or None
+
+        if event_type == "response.part.delta":
+            delta = data.get("delta")
+            if data.get("part_kind") == "text" and isinstance(delta, str):
+                text_deltas.append(delta)
+        elif event_type in ("response.part.created", "response.part.completed"):
+            part = data.get("part")
+            if isinstance(part, dict):
+                parts.append({"event": event_type, "part": part})
+                content = part.get("content")
+                if (
+                    event_type == "response.part.completed"
+                    and part.get("part_kind") == "text"
+                    and isinstance(content, dict)
+                    and isinstance(content.get("content"), str)
+                ):
+                    final_text = content["content"]
+        elif event_type in ("response.interaction.created", "response.interaction.resolved"):
+            interactions.append(data)
+        elif event_type == "response.chat.title.updated":
+            name = data.get("name")
+            if isinstance(name, str):
+                updated_title = name
+        elif event_type == "response.failed":
+            message = data.get("message") or "Chat V2 stream failed"
+            raise TransportError(
+                str(message),
+                {
+                    "code": data.get("code"),
+                    "response_id": response_id,
+                    "conversation_group_id": conversation_group_id,
+                    "outcome_uncertain": True,
+                    "events": raw_events,
+                },
+            )
+        elif event_type == "response.completed":
+            completed = True
+
+    if not completed:
+        raise TransportError(
+            "Chat V2 stream ended before response.completed. Inspect the history before retrying.",
+            {
+                "response_id": response_id,
+                "conversation_group_id": conversation_group_id,
+                "outcome_uncertain": True,
+                "events": raw_events,
+            },
+        )
+
+    return {
+        "stream": True,
+        "completed": True,
+        "response_id": response_id,
+        "conversation_group_id": conversation_group_id,
+        "final_text": final_text if final_text is not None else "".join(text_deltas),
+        "updated_title": updated_title,
+        "parts": parts,
+        "interactions": interactions,
+        "events": raw_events,
+    }
+
+
+def list_messages(
+    client: CodeerClient,
+    chat_id: int,
+    *,
+    external_user_id: Optional[str] = None,
+    limit: int = 500,
+) -> dict:
+    """Return every client-visible Chat V2 part for a persisted chat.
+
+    ``limit`` is the per-request page size, not a cap on the returned history.
+    The HTTP client unwraps Codeer's response envelope (and therefore its
+    pagination object), so exhaustion is detected from the number of messages
+    returned by each page. A final empty request is possible when the total is
+    an exact multiple of ``limit``.
+    """
+    if limit <= 0:
+        raise ValueError("limit must be greater than zero")
+
+    offset = 0
+    result: dict[str, Any] | None = None
+    messages: list[dict] = []
+    while True:
+        params: dict[str, Any] = {"limit": limit, "offset": offset}
+        if external_user_id is not None:
+            params["external_user_id"] = external_user_id
+        page = client.get(
+            f"/chats/{chat_id}/messages",
+            api_version="v2",
+            params=params,
+        )
+        if not isinstance(page, dict):
+            raise ValueError("Chat V2 messages response must be an object")
+
+        page_messages = page.get("messages") or []
+        if not isinstance(page_messages, list):
+            raise ValueError("Chat V2 messages response must contain a messages list")
+        if result is None:
+            result = dict(page)
+        else:
+            for key, value in page.items():
+                if key != "messages" and value is not None:
+                    result[key] = value
+        messages.extend(page_messages)
+
+        if len(page_messages) < limit:
+            break
+        offset += len(page_messages)
+
+    assert result is not None
+    result["messages"] = messages
+    return result
+
+
+def list_chats(client: CodeerClient) -> list[dict]:
+    """List chats through the legacy v1 endpoint; Chat V2 has no list route."""
+    return client.get("/chats")

{codeer_cli-0.1.10 → codeer_cli-0.1.12}/src/codeer_cli/client.py

@@ -12,11 +12,12 @@ from __future__ import annotations
 import json as json_lib
 import os
 from dataclasses import dataclass
-from typing import Any, Iterable, Iterator, Mapping, Optional
+from typing import Any, Iterable, Iterator, Literal, Mapping, Optional
 
 import httpx
 
 DEFAULT_CODEER_API_BASE = "https://api.codeer.ai"
+ApiVersion = Literal["v1", "v2"]
 
 
 class CodeerError(RuntimeError):
@@ -154,13 +155,14 @@ class CodeerClient:
         method: str,
         path: str,
         *,
+        api_version: ApiVersion = "v1",
         params: Optional[Mapping[str, Any]] = None,
         json: Any = None,
         files: Any = None,
         data: Any = None,
         timeout: Optional[float] = None,
     ) -> Any:
-        url = path if path.startswith("http") else f"/api/v1{path if path.startswith('/') else '/' + path}"
+        url = _api_url(path, api_version=api_version)
         request_kwargs: dict[str, Any] = {}
         if timeout is not None:
             request_kwargs["timeout"] = timeout
@@ -206,17 +208,28 @@ class CodeerClient:
         method: str,
         path: str,
         *,
+        api_version: ApiVersion = "v1",
         params: Optional[Mapping[str, Any]] = None,
         json: Any = None,
+        timeout: Optional[float] = None,
     ) -> Iterator[dict]:
         """Yield parsed SSE events from a streaming endpoint (e.g. POST /chats/{id}/messages).
 
         Each event is a dict like ``{"event": "message", "data": <parsed-json-or-str>}``.
         """
-        url = path if path.startswith("http") else f"/api/v1{path if path.startswith('/') else '/' + path}"
+        url = _api_url(path, api_version=api_version)
         method_upper = method.upper()
+        request_kwargs: dict[str, Any] = {}
+        if timeout is not None:
+            request_kwargs["timeout"] = timeout
         try:
-            with self._client.stream(method_upper, url, params=params, json=json) as r:
+            with self._client.stream(
+                method_upper,
+                url,
+                params=params,
+                json=json,
+                **request_kwargs,
+            ) as r:
                 if r.status_code >= 400:
                     body = r.read().decode("utf-8", "replace")
                     self._raise_for_error(r.status_code, body)
@@ -244,7 +257,7 @@ class CodeerClient:
                 method_upper,
                 path,
                 exc,
-                timeout_seconds=self.timeout,
+                timeout_seconds=timeout if timeout is not None else self.timeout,
             ) from exc
         except httpx.RequestError as exc:
             raise self._transport_error(method_upper, path, exc) from exc
@@ -314,6 +327,18 @@ def _maybe_json(raw: str) -> Any:
         return raw
 
 
+def _api_url(path: str, *, api_version: ApiVersion = "v1") -> str:
+    """Resolve a relative domain path without forcing every API onto one version."""
+    if path.startswith("http"):
+        return path
+    normalized = path if path.startswith("/") else f"/{path}"
+    if normalized.startswith("/api/"):
+        return normalized
+    if api_version not in ("v1", "v2"):
+        raise ValueError(f"Unsupported API version: {api_version}")
+    return f"/api/{api_version}{normalized}"
+
+
 def _workspace_names(profile: Mapping[str, Any]) -> dict[str, str]:
     names: dict[str, str] = {}
     for ws in profile.get("workspaces") or []: