pyrepowiki-cli 0.1.0__tar.gz → 0.2.0__tar.gz

{pyrepowiki_cli-0.1.0 → pyrepowiki_cli-0.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: pyrepowiki-cli
-Version: 0.1.0
+Version: 0.2.0
 Summary: CLI tool for querying DeepWiki/Google code wiki/Zread wiki documentation
 License: MIT
 Requires-Python: >=3.10
@@ -679,7 +679,7 @@ read commands need **no auth**. The commands `ask` and `submit` require a token
 ```bash
 repowiki-cli zread structure REPO [--lang zh|en] [--json]
 repowiki-cli zread contents REPO [SLUG] [--file PATH] [--start N] [--end M] [--lang zh|en] [--rich] [--json]
-repowiki-cli zread ask REPO [QUESTION] [--model MODEL] [--lang zh|en] [--rich] [--json] [--save PATH]
+repowiki-cli zread ask REPO [QUESTION] [--model MODEL] [--lang zh|en] [--rich] [--json] [--save PATH] [--stream] [--show-reasoning]
 repowiki-cli zread find QUERY [--limit N] [--lang zh|en] [--json]
 repowiki-cli zread stat REPO [--lang zh|en] [--human] [--stale] [--json]
 repowiki-cli zread search REPO QUERY [--lang zh|en] [--json]
@@ -730,12 +730,15 @@ repowiki-cli zread contents https://github.com/o/r/blob/main/src/a.py#L10-L20
 ### `zread ask`
 
 ```bash
-repowiki-cli zread ask REPO [QUESTION] [--model MODEL] [--rich] [--json] [--save PATH]
+repowiki-cli zread ask REPO [QUESTION] [--model MODEL] [--rich] [--json] [--save PATH] [--stream] [--show-reasoning]
 ```
 
 With `QUESTION`, `ask` answers once and exits. Without it, `ask` starts an
 interactive REPL — type one question per line and `/exit` (or `/quit`/`/q`) to
-quit.
+quit. In the REPL each turn is **threaded**: the same `talk_id` is reused and the
+previous answer's message id is sent as `parent_message_id`, so follow-up
+questions carry conversational memory. Type `/new` (or `/reset`) to start a fresh
+thread.
 
 Unlike the other commands, `ask` requires an auth token (a JWT). Log in to
 zread.ai, open the browser DevTools console, and run
@@ -746,9 +749,17 @@ the token. Set it as the `ZREAD_TOKEN` environment variable.
   variable). This maps to the web UI's `CGX_CHAT_MODEL` (e.g. `glm-5.1`,
   `claude-sonnet-4.6`).
 - `--lang zh|en` — language.
-- `--rich` — render the answer's Markdown with `rich`.
-- `--json` — emit a JSON envelope. Ignored in interactive mode.
+- `--rich` — render the answer's Markdown with `rich`. Has no effect with
+  `--stream` (the answer streams as plain text).
+- `--json` — emit a JSON envelope. Ignored in interactive mode. Has no effect
+  with `--stream`.
 - `--save PATH` — save the answer to a Markdown file (see *Saving*).
+- `--stream` — stream the answer's `answer` SSE chunks to stdout as they arrive
+  (plain text), instead of buffering until the full answer is ready.
+- `--show-reasoning` — also show the model's reasoning trace (the
+  `reasoning_content` SSE field, streamed before the answer). In `--stream` mode
+  it prints a `reasoning:` block before the `answer:` block; otherwise it prints
+  the trace dimmed above the answer. Has no effect with `--json`.
 
 ### `zread find`
 

{pyrepowiki_cli-0.1.0 → pyrepowiki_cli-0.2.0}/README.md

@@ -666,7 +666,7 @@ read commands need **no auth**. The commands `ask` and `submit` require a token
 ```bash
 repowiki-cli zread structure REPO [--lang zh|en] [--json]
 repowiki-cli zread contents REPO [SLUG] [--file PATH] [--start N] [--end M] [--lang zh|en] [--rich] [--json]
-repowiki-cli zread ask REPO [QUESTION] [--model MODEL] [--lang zh|en] [--rich] [--json] [--save PATH]
+repowiki-cli zread ask REPO [QUESTION] [--model MODEL] [--lang zh|en] [--rich] [--json] [--save PATH] [--stream] [--show-reasoning]
 repowiki-cli zread find QUERY [--limit N] [--lang zh|en] [--json]
 repowiki-cli zread stat REPO [--lang zh|en] [--human] [--stale] [--json]
 repowiki-cli zread search REPO QUERY [--lang zh|en] [--json]
@@ -717,12 +717,15 @@ repowiki-cli zread contents https://github.com/o/r/blob/main/src/a.py#L10-L20
 ### `zread ask`
 
 ```bash
-repowiki-cli zread ask REPO [QUESTION] [--model MODEL] [--rich] [--json] [--save PATH]
+repowiki-cli zread ask REPO [QUESTION] [--model MODEL] [--rich] [--json] [--save PATH] [--stream] [--show-reasoning]
 ```
 
 With `QUESTION`, `ask` answers once and exits. Without it, `ask` starts an
 interactive REPL — type one question per line and `/exit` (or `/quit`/`/q`) to
-quit.
+quit. In the REPL each turn is **threaded**: the same `talk_id` is reused and the
+previous answer's message id is sent as `parent_message_id`, so follow-up
+questions carry conversational memory. Type `/new` (or `/reset`) to start a fresh
+thread.
 
 Unlike the other commands, `ask` requires an auth token (a JWT). Log in to
 zread.ai, open the browser DevTools console, and run
@@ -733,9 +736,17 @@ the token. Set it as the `ZREAD_TOKEN` environment variable.
   variable). This maps to the web UI's `CGX_CHAT_MODEL` (e.g. `glm-5.1`,
   `claude-sonnet-4.6`).
 - `--lang zh|en` — language.
-- `--rich` — render the answer's Markdown with `rich`.
-- `--json` — emit a JSON envelope. Ignored in interactive mode.
+- `--rich` — render the answer's Markdown with `rich`. Has no effect with
+  `--stream` (the answer streams as plain text).
+- `--json` — emit a JSON envelope. Ignored in interactive mode. Has no effect
+  with `--stream`.
 - `--save PATH` — save the answer to a Markdown file (see *Saving*).
+- `--stream` — stream the answer's `answer` SSE chunks to stdout as they arrive
+  (plain text), instead of buffering until the full answer is ready.
+- `--show-reasoning` — also show the model's reasoning trace (the
+  `reasoning_content` SSE field, streamed before the answer). In `--stream` mode
+  it prints a `reasoning:` block before the `answer:` block; otherwise it prints
+  the trace dimmed above the answer. Has no effect with `--json`.
 
 ### `zread find`
 

{pyrepowiki_cli-0.1.0 → pyrepowiki_cli-0.2.0}/README.zh-CN.md

@@ -621,7 +621,7 @@ repowiki-cli codewiki cp REPO [OUTPUT_DIR]
 ```bash
 repowiki-cli zread structure REPO [--lang zh|en] [--json]
 repowiki-cli zread contents REPO [SLUG] [--file PATH] [--start N] [--end M] [--lang zh|en] [--rich] [--json]
-repowiki-cli zread ask REPO [QUESTION] [--model MODEL] [--lang zh|en] [--rich] [--json] [--save PATH]
+repowiki-cli zread ask REPO [QUESTION] [--model MODEL] [--lang zh|en] [--rich] [--json] [--save PATH] [--stream] [--show-reasoning]
 repowiki-cli zread find QUERY [--limit N] [--lang zh|en] [--json]
 repowiki-cli zread stat REPO [--lang zh|en] [--human] [--stale] [--json]
 repowiki-cli zread search REPO QUERY [--lang zh|en] [--json]
@@ -668,11 +668,13 @@ repowiki-cli zread contents https://github.com/o/r/blob/main/src/a.py#L10-L20
 ### `zread ask`
 
 ```bash
-repowiki-cli zread ask REPO [QUESTION] [--model MODEL] [--rich] [--json] [--save PATH]
+repowiki-cli zread ask REPO [QUESTION] [--model MODEL] [--rich] [--json] [--save PATH] [--stream] [--show-reasoning]
 ```
 
 带 `QUESTION` 则单次回答后退出；不带则进入交互式 REPL——每行一个问题，输入
-`/exit`（或 `/quit`/`/q`）退出。
+`/exit`（或 `/quit`/`/q`）退出。REPL 中每轮都是**多轮续接**的：复用同一个
+`talk_id`，并把上一轮答案的 message id 作为 `parent_message_id` 发送，因此追问
+能携带上下文记忆。输入 `/new`（或 `/reset`）可开启一个新线程。
 
 与其他命令不同，`ask` 需要鉴权 token（一个 JWT）。登录 zread.ai 后，在浏览器
 DevTools 控制台执行
@@ -682,9 +684,13 @@ DevTools 控制台执行
 - `--model MODEL` — 模型（默认 `glm-5.1`，或环境变量 `ZREAD_MODEL`）。对应网页端的
   `CGX_CHAT_MODEL`（如 `glm-5.1`、`claude-sonnet-4.6`）。
 - `--lang zh|en` — 语言。
-- `--rich` — 用 `rich` 渲染答案的 Markdown。
-- `--json` — 输出 JSON 信封；交互模式下忽略。
+- `--rich` — 用 `rich` 渲染答案的 Markdown。与 `--stream` 一起使用时无效（流式输出为纯文本）。
+- `--json` — 输出 JSON 信封；交互模式下忽略。与 `--stream` 一起使用时无效。
 - `--save PATH` — 把答案保存为 Markdown 文件（见「保存」）。
+- `--stream` — 通过 SSE 把 `answer` 事件的分片逐字写到 stdout，流式输出纯文本答案。
+- `--show-reasoning` — 同时展示模型的思考过程（SSE 的 `reasoning_content` 字段，
+  在回答之前流式下发）。`--stream` 模式下会先输出 `reasoning:` 块再输出
+  `answer:` 块；否则以暗色打印在答案上方。与 `--json` 一起使用时无效。
 
 ### `zread find`
 

{pyrepowiki_cli-0.1.0 → pyrepowiki_cli-0.2.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "pyrepowiki-cli"
-version = "0.1.0"
+version = "0.2.0"
 description = "CLI tool for querying DeepWiki/Google code wiki/Zread wiki documentation"
 readme = "README.md"
 requires-python = ">=3.10"

{pyrepowiki_cli-0.1.0 → pyrepowiki_cli-0.2.0}/src/repowiki/services/zread/cli.py

@@ -137,6 +137,49 @@ def _repl_prompt() -> str:
     return typer.style(">> ", fg=typer.colors.CYAN, bold=True)
 
 
+def _stream_chunk(text: str) -> None:
+    sys.stdout.write(text)
+    sys.stdout.flush()
+
+
+def _streamers(show_reasoning: bool):
+    """Return ``(on_reasoning, on_chunk)`` callbacks for ``--stream``.
+
+    Without ``--show-reasoning`` this streams only the answer (reasoning is
+    dropped). With it, the reasoning trace streams first under a ``reasoning:``
+    label, then the answer under an ``answer:`` label.
+    """
+    if not show_reasoning:
+        return None, _stream_chunk
+    state = {"reasoning": False, "answer": False}
+
+    def on_reasoning(text: str) -> None:
+        if not state["reasoning"]:
+            sys.stdout.write("\nreasoning:\n")
+            state["reasoning"] = True
+        sys.stdout.write(text)
+        sys.stdout.flush()
+
+    def on_chunk(text: str) -> None:
+        if not state["answer"]:
+            if state["reasoning"]:
+                sys.stdout.write("\n\nanswer:\n")
+            state["answer"] = True
+        sys.stdout.write(text)
+        sys.stdout.flush()
+
+    return on_reasoning, on_chunk
+
+
+def _print_reasoning(reasoning: str) -> None:
+    """Print a captured reasoning trace (dimmed) above the answer."""
+    if not reasoning.strip():
+        return
+    typer.secho("reasoning:", fg=typer.colors.YELLOW, dim=True)
+    typer.secho(reasoning.strip(), fg=typer.colors.YELLOW, dim=True)
+    typer.echo()
+
+
 def _print_error(exc: Exception) -> None:
     if isinstance(exc, ZreadError):
         message = str(exc)
@@ -155,10 +198,16 @@ def _append_save(path: Optional[str], repo: str, question: str, answer: str) ->
 
 
 async def _repl(
-    resolved: str, rich: bool, save_path: Optional[str], model: Optional[str], lang: Optional[str]
+    resolved: str, rich: bool, save_path: Optional[str], model: Optional[str], lang: Optional[str],
+    stream: bool = False, show_reasoning: bool = False,
 ) -> None:
     prompt = _repl_prompt()
     client = _make_client(lang, model)
+    try:
+        talk = await client.start_talk(resolved)
+    except Exception as exc:
+        _print_error(exc)
+        return
     while True:
         try:
             line = input(prompt)
@@ -170,19 +219,37 @@ async def _repl(
             continue
         if q in ("/exit", "/quit", "/q"):
             break
+        if q in ("/new", "/reset"):
+            try:
+                talk = await client.start_talk(resolved)
+                typer.echo("Started a new thread.")
+            except Exception as exc:
+                _print_error(exc)
+            continue
         try:
-            with status("Thinking..."):
-                answer = await client.ask(resolved, q)
+            if stream:
+                on_reasoning, on_chunk = _streamers(show_reasoning)
+                answer = await talk.ask(q, on_chunk=on_chunk, on_reasoning=on_reasoning)
+            else:
+                reasoning_parts: list[str] = []
+                with status("Thinking..."):
+                    answer = await talk.ask(
+                        q, on_reasoning=reasoning_parts.append if show_reasoning else None
+                    )
         except Exception as exc:
             _print_error(exc)
             continue
         _append_save(save_path, resolved, q, answer)
-        typer.echo()
-        if rich:
-            render_rich(answer.strip())
+        if stream:
+            typer.echo()
         else:
-            typer.echo(answer.strip())
-        typer.echo()
+            _print_reasoning("".join(reasoning_parts))
+            typer.echo()
+            if rich:
+                render_rich(answer.strip())
+            else:
+                typer.echo(answer.strip())
+            typer.echo()
 
 
 def _render_structure(pages: list[Page]) -> str:
@@ -416,24 +483,65 @@ def ask(
     rich: bool = typer.Option(False, "--rich", help="Render Markdown with rich"),
     json: bool = typer.Option(False, "--json", help="Emit machine-readable JSON"),
     save: Optional[str] = typer.Option(None, "--save", help="Save the answer to a Markdown file."),
+    stream: bool = typer.Option(False, "--stream", help="Stream answer chunks as they arrive"),
+    show_reasoning: bool = typer.Option(
+        False, "--show-reasoning", help="Also show the model's reasoning trace",
+    ),
 ) -> None:
     """Ask a question about a repository (single-shot or interactive)."""
     resolved = _resolve_repo(repo, json)
     if question is not None:
         if not question.strip():
             _fail("Question must not be empty.", "invalid_input", json)
+        if stream and json:
+            typer.secho(
+                "Warning: --stream has no effect with --json.",
+                fg=typer.colors.YELLOW, err=True,
+            )
+        if stream and rich:
+            typer.secho(
+                "Warning: --rich has no effect with --stream; the answer streams as plain text.",
+                fg=typer.colors.YELLOW, err=True,
+            )
+        if show_reasoning and json:
+            typer.secho(
+                "Warning: --show-reasoning has no effect with --json.",
+                fg=typer.colors.YELLOW, err=True,
+            )
         if (mock := _mock_text()) is not None:
             answer = Answer(body=mock)
-        else:
-            client = _make_client(lang, model)
-            try:
+            if json:
+                typer.echo(format_json(resolved, "ask", question=question, answer=answer.body))
+            else:
+                _emit(resolved, "ask", format_answer(answer), rich, False)
+            _append_save(save, resolved, question, answer.body)
+            return
+        client = _make_client(lang, model)
+        streamed = False
+        reasoning_parts: list[str] = []
+        try:
+            if stream and not json:
+                on_reasoning, on_chunk = _streamers(show_reasoning)
+                answer = Answer(body=run_async(
+                    client.ask(resolved, question, on_chunk=on_chunk, on_reasoning=on_reasoning)
+                ))
+                streamed = True
+            else:
                 with status("Thinking..."):
-                    answer = Answer(body=run_async(client.ask(resolved, question)))
-            except Exception as exc:
-                _handle_exception(exc, json)
+                    answer = Answer(body=run_async(
+                        client.ask(
+                            resolved, question,
+                            on_reasoning=reasoning_parts.append if show_reasoning else None,
+                        )
+                    ))
+        except Exception as exc:
+            _handle_exception(exc, json)
         if json:
             typer.echo(format_json(resolved, "ask", question=question, answer=answer.body))
+        elif streamed:
+            typer.echo()
         else:
+            _print_reasoning("".join(reasoning_parts))
             _emit(resolved, "ask", format_answer(answer), rich, False)
         _append_save(save, resolved, question, answer.body)
         return
@@ -441,10 +549,10 @@ def ask(
         typer.secho("Warning: --json has no effect in interactive mode.", fg=typer.colors.YELLOW, err=True)
     typer.echo(format_header("Zread", resolved, "ask"))
     typer.echo()
-    typer.echo("Ask a question, or /exit to quit.")
+    typer.echo("Ask a question, or /exit to quit.  (/new starts a new thread)")
     typer.echo()
     try:
-        run_async(_repl(resolved, rich, save, model, lang))
+        run_async(_repl(resolved, rich, save, model, lang, stream, show_reasoning))
     except Exception as exc:
         _handle_exception(exc, False)
 

{pyrepowiki_cli-0.1.0 → pyrepowiki_cli-0.2.0}/src/repowiki/services/zread/client.py

@@ -84,31 +84,76 @@ def _unwrap(resp: httpx.Response) -> object:
     return result.get("data")
 
 
-def _parse_sse_body(lines: list[str]) -> str:
-    event = None
-    finished: list[str] = []
-    chunks: list[str] = []
-    for line in lines:
+class _SSEParser:
+    """Incrementally parse the talk SSE stream.
+
+    ``answer`` events carry incremental chunks — ``data.text`` is the answer and
+    ``data.reasoning_content`` is the model's thinking trace (streamed first, then
+    empty during the answer phase). ``round_finish`` carries the complete text,
+    ``finish`` ends the stream. ``on_chunk`` / ``on_reasoning``, when given, are
+    invoked with each ``answer`` chunk as it arrives.
+    """
+
+    def __init__(self, on_chunk=None, on_reasoning=None) -> None:
+        self._on_chunk = on_chunk
+        self._on_reasoning = on_reasoning
+        self._event: str | None = None
+        self._finished: list[str] = []
+        self._chunks: list[str] = []
+        self._reasoning: list[str] = []
+        self._message_id = ""
+
+    def feed(self, line: str) -> bool:
+        """Consume one raw SSE line; return True when the stream is done."""
         line = line.strip()
         if not line:
-            continue
+            return False
         if line.startswith("event:"):
-            event = line[6:].strip()
-            if event == "finish":
-                break
-            if event == "error":
+            self._event = line[6:].strip()
+            if self._event == "finish":
+                return True
+            if self._event == "error":
                 raise ZreadError("talk stream returned an error event")
         elif line.startswith("data:"):
             try:
                 data = json.loads(line[5:].strip())
             except json.JSONDecodeError:
-                continue
-            if event == "round_finish":
-                finished.append(data.get("text", ""))
-            elif event == "answer":
-                chunks.append(data.get("text", ""))
-    text = "\n".join(finished) or "".join(chunks)
-    return text.strip()
+                return False
+            if data.get("id"):
+                self._message_id = data["id"]
+            if self._event == "round_finish":
+                self._finished.append(data.get("text", ""))
+            elif self._event == "answer":
+                reasoning = data.get("reasoning_content", "")
+                if reasoning:
+                    self._reasoning.append(reasoning)
+                    if self._on_reasoning is not None:
+                        self._on_reasoning(reasoning)
+                text = data.get("text", "")
+                self._chunks.append(text)
+                if self._on_chunk is not None and text:
+                    self._on_chunk(text)
+        return False
+
+    @property
+    def text(self) -> str:
+        return ("\n".join(self._finished) or "".join(self._chunks)).strip()
+
+    @property
+    def reasoning(self) -> str:
+        return "".join(self._reasoning).strip()
+
+    @property
+    def message_id(self) -> str:
+        return self._message_id
+
+
+def _parse_sse_body(lines: list[str], on_chunk=None, on_reasoning=None) -> str:
+    parser = _SSEParser(on_chunk, on_reasoning)
+    for line in lines:
+        if parser.feed(line):
+            break
+    return parser.text
 
 
 class ZreadClient:
@@ -349,7 +394,18 @@ class ZreadClient:
             raise ZreadNotFoundError(f"no wiki_id for {repo}")
         return wiki_id
 
-    async def ask(self, repo: str, question: str) -> str:
+    async def ask(self, repo: str, question: str, on_chunk=None, on_reasoning=None) -> str:
+        if not self._token:
+            raise ZreadError(
+                "ZREAD_TOKEN is required for ask. Get it from zread.ai after login: "
+                "run JSON.parse(localStorage.getItem('CGX_AUTH_STORAGE')).state.token in "
+                "the DevTools console, then export ZREAD_TOKEN."
+            )
+        talk = await self.start_talk(repo)
+        return await talk.ask(question, on_chunk=on_chunk, on_reasoning=on_reasoning)
+
+    async def start_talk(self, repo: str) -> "ZreadTalk":
+        """Resolve a repo and open a new talk thread (``talk_id``)."""
         if not self._token:
             raise ZreadError(
                 "ZREAD_TOKEN is required for ask. Get it from zread.ai after login: "
@@ -365,7 +421,7 @@ class ZreadClient:
         if not pages:
             raise ZreadNotFoundError(f"{repo} has no wiki pages")
         talk_id = await self._create_talk()
-        return await self._send_message(talk_id, question, wiki_id, pages[0].page_id, repo_id)
+        return ZreadTalk(self, repo_id, wiki_id, pages[0].page_id, talk_id)
 
     async def _create_talk(self) -> str:
         resp = await self._request(
@@ -382,7 +438,8 @@ class ZreadClient:
     async def _send_message(
         self, talk_id: str, question: str,
         wiki_id: str, page_id: str, repo_id: str,
-    ) -> str:
+        on_chunk=None, on_reasoning=None, parent_message_id: str = "",
+    ) -> tuple[str, str]:
         url = f"{BASE}/api/v1/talk/{talk_id}/message"
         headers = self._headers({
             "Authorization": f"Bearer {self._token}",
@@ -392,7 +449,7 @@ class ZreadClient:
         if repo_id:
             context["repo"] = {"repo_id": repo_id}
         body = {
-            "parent_message_id": "",
+            "parent_message_id": parent_message_id,
             "query": question,
             "context": context,
             "model": self._model,
@@ -410,5 +467,37 @@ class ZreadClient:
                         f"Zread returned HTTP {resp.status_code} for {url}"
                         + (f": {detail}" if detail else "")
                     )
-                lines = [line async for line in resp.aiter_lines()]
-        return _parse_sse_body(lines)
+                parser = _SSEParser(on_chunk, on_reasoning)
+                async for line in resp.aiter_lines():
+                    if parser.feed(line):
+                        break
+        return parser.text, parser.message_id
+
+
+class ZreadTalk:
+    """A single threaded conversation against one repo.
+
+    Holds the ``talk_id`` and chains messages via ``parent_message_id`` (the id
+    of the previous assistant message, captured from the SSE ``id`` field) so
+    follow-up questions carry conversational memory.
+    """
+
+    def __init__(
+        self, client: "ZreadClient", repo_id: str, wiki_id: str, page_id: str, talk_id: str
+    ) -> None:
+        self._client = client
+        self._repo_id = repo_id
+        self._wiki_id = wiki_id
+        self._page_id = page_id
+        self._talk_id = talk_id
+        self._parent_message_id = ""
+
+    async def ask(self, question: str, on_chunk=None, on_reasoning=None) -> str:
+        answer, message_id = await self._client._send_message(
+            self._talk_id, question, self._wiki_id, self._page_id, self._repo_id,
+            on_chunk=on_chunk, on_reasoning=on_reasoning,
+            parent_message_id=self._parent_message_id,
+        )
+        if message_id:
+            self._parent_message_id = message_id
+        return answer

{pyrepowiki_cli-0.1.0 → pyrepowiki_cli-0.2.0}/tests/test_zread_cli.py

@@ -156,12 +156,16 @@ def test_ask_repl(monkeypatch):
     inputs = iter(["What is Fiber?", "/exit"])
     monkeypatch.setattr("builtins.input", lambda prompt="": next(inputs))
 
+    class FakeTalk:
+        async def ask(self, question, on_chunk=None, on_reasoning=None):
+            return f"answer to {question}"
+
     class FakeClient:
         def __init__(self, **kwargs):
             pass
 
-        async def ask(self, repo, question):
-            return f"answer to {question}"
+        async def start_talk(self, repo):
+            return FakeTalk()
 
     monkeypatch.setattr("repowiki.services.zread.cli.ZreadClient", FakeClient)
     result = runner.invoke(zread_app, ["ask", "facebook/react"])
@@ -169,6 +173,138 @@ def test_ask_repl(monkeypatch):
     assert "answer to What is Fiber?" in result.output
 
 
+def test_ask_repl_reuses_talk(monkeypatch):
+    inputs = iter(["q1", "q2", "/exit"])
+    monkeypatch.setattr("builtins.input", lambda prompt="": next(inputs))
+    started = {"n": 0}
+
+    class FakeTalk:
+        async def ask(self, question, on_chunk=None, on_reasoning=None):
+            return f"answer to {question}"
+
+    class FakeClient:
+        def __init__(self, **kwargs):
+            pass
+
+        async def start_talk(self, repo):
+            started["n"] += 1
+            return FakeTalk()
+
+    monkeypatch.setattr("repowiki.services.zread.cli.ZreadClient", FakeClient)
+    result = runner.invoke(zread_app, ["ask", "facebook/react"])
+    assert result.exit_code == 0
+    assert "answer to q1" in result.output
+    assert "answer to q2" in result.output
+    assert started["n"] == 1
+
+
+def test_ask_repl_new_thread(monkeypatch):
+    inputs = iter(["q1", "/new", "/exit"])
+    monkeypatch.setattr("builtins.input", lambda prompt="": next(inputs))
+    started = {"n": 0}
+
+    class FakeTalk:
+        async def ask(self, question, on_chunk=None, on_reasoning=None):
+            return f"answer to {question}"
+
+    class FakeClient:
+        def __init__(self, **kwargs):
+            pass
+
+        async def start_talk(self, repo):
+            started["n"] += 1
+            return FakeTalk()
+
+    monkeypatch.setattr("repowiki.services.zread.cli.ZreadClient", FakeClient)
+    result = runner.invoke(zread_app, ["ask", "facebook/react"])
+    assert result.exit_code == 0
+    assert started["n"] == 2
+
+
+def test_ask_stream_emits_chunks(monkeypatch):
+    class FakeClient:
+        def __init__(self, **kwargs):
+            pass
+
+        async def ask(self, repo, question, on_chunk=None, on_reasoning=None):
+            if on_chunk is not None:
+                on_chunk("hel")
+                on_chunk("lo")
+            return "hello"
+
+    monkeypatch.setattr("repowiki.services.zread.cli.ZreadClient", FakeClient)
+    result = runner.invoke(zread_app, ["ask", "owner/example", "q?", "--stream"])
+    assert result.exit_code == 0
+    assert "hello" in result.output
+
+
+def test_ask_stream_warns_json(monkeypatch):
+    class FakeClient:
+        def __init__(self, **kwargs):
+            pass
+
+        async def ask(self, repo, question, on_chunk=None, on_reasoning=None):
+            return "hello"
+
+    monkeypatch.setattr("repowiki.services.zread.cli.ZreadClient", FakeClient)
+    result = runner.invoke(zread_app, ["ask", "owner/example", "q?", "--stream", "--json"])
+    assert result.exit_code == 0
+    assert "--stream has no effect with --json" in result.output
+
+
+def test_ask_show_reasoning(monkeypatch):
+    class FakeClient:
+        def __init__(self, **kwargs):
+            pass
+
+        async def ask(self, repo, question, on_chunk=None, on_reasoning=None):
+            if on_reasoning is not None:
+                on_reasoning("think ")
+                on_reasoning("hard")
+            return "hello"
+
+    monkeypatch.setattr("repowiki.services.zread.cli.ZreadClient", FakeClient)
+    result = runner.invoke(zread_app, ["ask", "owner/example", "q?", "--show-reasoning"])
+    assert result.exit_code == 0
+    assert "think hard" in result.output
+    assert "hello" in result.output
+
+
+def test_ask_show_reasoning_warns_json(monkeypatch):
+    class FakeClient:
+        def __init__(self, **kwargs):
+            pass
+
+        async def ask(self, repo, question, on_chunk=None, on_reasoning=None):
+            return "hello"
+
+    monkeypatch.setattr("repowiki.services.zread.cli.ZreadClient", FakeClient)
+    result = runner.invoke(zread_app, ["ask", "owner/example", "q?", "--show-reasoning", "--json"])
+    assert result.exit_code == 0
+    assert "--show-reasoning has no effect with --json" in result.output
+
+
+def test_ask_stream_show_reasoning(monkeypatch):
+    class FakeClient:
+        def __init__(self, **kwargs):
+            pass
+
+        async def ask(self, repo, question, on_chunk=None, on_reasoning=None):
+            if on_reasoning is not None:
+                on_reasoning("think")
+            if on_chunk is not None:
+                on_chunk("hello")
+            return "hello"
+
+    monkeypatch.setattr("repowiki.services.zread.cli.ZreadClient", FakeClient)
+    result = runner.invoke(zread_app, ["ask", "owner/example", "q?", "--stream", "--show-reasoning"])
+    assert result.exit_code == 0
+    assert "reasoning:" in result.output
+    assert "answer:" in result.output
+    assert "think" in result.output
+    assert "hello" in result.output
+
+
 def test_cp_concurrency_zero_fails(monkeypatch):
     monkeypatch.setenv("REPOWIKI_ZREAD_MOCK", "- Overview")
     result = runner.invoke(zread_app, ["cp", "owner/example", "--concurrency", "0"])

{pyrepowiki_cli-0.1.0 → pyrepowiki_cli-0.2.0}/tests/test_zread_client.py

@@ -297,6 +297,40 @@ def test_parse_sse_body_errors_on_error_event():
         _parse_sse_body(["event: error", "data: {}"])
 
 
+def test_parse_sse_body_streams_answer_chunks():
+    chunks = []
+    lines = [
+        "event: answer",
+        'data: {"text": "hel"}',
+        "event: answer",
+        'data: {"text": "lo"}',
+        "event: round_finish",
+        'data: {"text": "hello"}',
+        "event: finish",
+        "data: {}",
+    ]
+    assert _parse_sse_body(lines, on_chunk=chunks.append) == "hello"
+    assert chunks == ["hel", "lo"]
+
+
+def test_parse_sse_body_streams_reasoning():
+    reasoning = []
+    lines = [
+        "event: answer",
+        'data: {"text": "", "reasoning_content": "think "}',
+        "event: answer",
+        'data: {"text": "", "reasoning_content": "hard"}',
+        "event: answer",
+        'data: {"text": "ans", "reasoning_content": ""}',
+        "event: round_finish",
+        'data: {"text": "answer"}',
+        "event: finish",
+        "data: {}",
+    ]
+    assert _parse_sse_body(lines, on_reasoning=reasoning.append) == "answer"
+    assert reasoning == ["think ", "hard"]
+
+
 def test_ask_requires_token():
     client = ZreadClient(transport=httpx.MockTransport(lambda r: httpx.Response(200)), token=None)
     with pytest.raises(ZreadError, match="ZREAD_TOKEN"):
@@ -337,6 +371,67 @@ def test_ask_flow(monkeypatch):
     assert seen["body"]["context"]["wiki"] == {"page_id": "p0", "wiki_id": "w1"}
 
 
+def test_ask_streams_chunks(monkeypatch):
+    import repowiki.services.zread.client as mod
+
+    monkeypatch.setattr(mod.asyncio, "sleep", _no_sleep)
+
+    sse = (
+        'event: answer\ndata: {"text": "hel"}\n\n'
+        'event: answer\ndata: {"text": "lo"}\n\n'
+        'event: round_finish\ndata: {"text": "hello"}\n\n'
+        'event: finish\ndata: {}\n\n'
+    )
+
+    def handler(request):
+        url = str(request.url)
+        if url.endswith("/api/v1/talk"):
+            return httpx.Response(200, json={"code": 0, "data": {"talk_id": "t1"}})
+        if "/message" in url:
+            return httpx.Response(200, content=sse.encode(), headers={"content-type": "text/event-stream"})
+        if "repo/github" in url:
+            return httpx.Response(200, json={"code": 0, "data": {"repo_id": "r1", "wiki_id": "w1"}})
+        if "/api/v1/wiki/" in url:
+            return httpx.Response(200, json=_wiki_json(1))
+        return httpx.Response(404)
+
+    client = ZreadClient(transport=httpx.MockTransport(handler), token="tok")
+    chunks = []
+    answer = run_async(client.ask("o/r", "q?", on_chunk=chunks.append))
+    assert answer == "hello"
+    assert chunks == ["hel", "lo"]
+
+
+def test_talk_threads_messages(monkeypatch):
+    import repowiki.services.zread.client as mod
+
+    monkeypatch.setattr(mod.asyncio, "sleep", _no_sleep)
+    parents = []
+
+    def handler(request):
+        url = str(request.url)
+        if url.endswith("/api/v1/talk"):
+            return httpx.Response(200, json={"code": 0, "data": {"talk_id": "t1"}})
+        if "/message" in url:
+            parents.append(json.loads(request.content)["parent_message_id"])
+            sse = (
+                'event: round_finish\ndata: {"id": "m1", "text": "first"}\n\n'
+                'event: finish\ndata: {"id": "m1"}\n\n'
+            )
+            return httpx.Response(200, content=sse.encode(), headers={"content-type": "text/event-stream"})
+        if "repo/github" in url:
+            return httpx.Response(200, json={"code": 0, "data": {"repo_id": "r1", "wiki_id": "w1"}})
+        if "/api/v1/wiki/" in url:
+            return httpx.Response(200, json=_wiki_json(1))
+        return httpx.Response(404)
+
+    client = ZreadClient(transport=httpx.MockTransport(handler), token="tok")
+    talk = run_async(client.start_talk("o/r"))
+    assert run_async(talk.ask("q1?")) == "first"
+    assert run_async(talk.ask("q2?")) == "first"
+    assert parents == ["", "m1"]
+
+
 def test_is_cert_error_detects_ssl_verification():
     from repowiki.services.zread.client import _is_cert_error
 

{pyrepowiki_cli-0.1.0 → pyrepowiki_cli-0.2.0}/uv.lock

@@ -600,6 +600,40 @@ crypto = [
     { name = "cryptography" },
 ]
 
+[[package]]
+name = "pyrepowiki-cli"
+version = "0.1.0"
+source = { editable = "." }
+dependencies = [
+    { name = "httpx" },
+    { name = "mcp" },
+    { name = "rich" },
+    { name = "typer" },
+    { name = "websockets", version = "16.1.1", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
+    { name = "websockets", version = "17.1", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },
+]
+
+[package.dev-dependencies]
+dev = [
+    { name = "pytest" },
+    { name = "pytest-asyncio" },
+]
+
+[package.metadata]
+requires-dist = [
+    { name = "httpx", specifier = ">=0.27.0" },
+    { name = "mcp", specifier = ">=1.0.0,<2.0.0" },
+    { name = "rich", specifier = ">=13.0.0" },
+    { name = "typer", specifier = ">=0.12.0" },
+    { name = "websockets", specifier = ">=13.0.0" },
+]
+
+[package.metadata.requires-dev]
+dev = [
+    { name = "pytest", specifier = ">=8.0.0" },
+    { name = "pytest-asyncio", specifier = ">=0.23.0" },
+]
+
 [[package]]
 name = "pytest"
 version = "9.1.1"
@@ -690,40 +724,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/2c/58/ca301544e1fa93ed4f80d724bf5b194f6e4b945841c5bfd555878eea9fcb/referencing-0.37.0-py3-none-any.whl", hash = "sha256:381329a9f99628c9069361716891d34ad94af76e461dcb0335825aecc7692231", size = 26766, upload-time = "2025-10-13T15:30:47.625Z" },
 ]
 
-[[package]]
-name = "repowiki-cli"
-version = "0.1.0"
-source = { editable = "." }
-dependencies = [
-    { name = "httpx" },
-    { name = "mcp" },
-    { name = "rich" },
-    { name = "typer" },
-    { name = "websockets", version = "16.1.1", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
-    { name = "websockets", version = "17.1", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },
-]
-
-[package.dev-dependencies]
-dev = [
-    { name = "pytest" },
-    { name = "pytest-asyncio" },
-]
-
-[package.metadata]
-requires-dist = [
-    { name = "httpx", specifier = ">=0.27.0" },
-    { name = "mcp", specifier = ">=1.0.0,<2.0.0" },
-    { name = "rich", specifier = ">=13.0.0" },
-    { name = "typer", specifier = ">=0.12.0" },
-    { name = "websockets", specifier = ">=13.0.0" },
-]
-
-[package.metadata.requires-dev]
-dev = [
-    { name = "pytest", specifier = ">=8.0.0" },
-    { name = "pytest-asyncio", specifier = ">=0.23.0" },
-]
-
 [[package]]
 name = "rich"
 version = "15.0.0"