anna-agent 0.2.0__tar.gz → 0.2.2__tar.gz

{anna_agent-0.2.0 → anna_agent-0.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: anna-agent
-Version: 0.2.0
+Version: 0.2.2
 Summary: Dynamic Evolution Agent System with Multi-Session Memory for Realistic Seeker Simulation
 Project-URL: Documentation, https://sci-m-wang.github.io/AnnaAgent/
 Project-URL: Repository, https://github.com/sci-m-wang/AnnaAgent
@@ -61,7 +61,7 @@ After installation, the short command `anna` is available in any terminal:
 
 ```bash
 anna --version
-anna init anna-workspace
+anna create anna-workspace
 anna doctor --workspace anna-workspace
 ```
 
@@ -72,7 +72,7 @@ a workspace deployment environment on the Linux/GPU machine. This keeps the
 lightweight `anna` CLI install separate from heavy vLLM dependencies:
 
 ```bash
-anna init anna-workspace --deploy-env
+anna create anna-workspace --deploy-env
 
 # Or add it later to an existing workspace.
 anna models env setup --workspace anna-workspace
@@ -117,17 +117,17 @@ First, create a workspace, choose whether to use the SFT modules, and let the CL
 write the resulting configuration:
 
 ```bash
-anna init anna-workspace
+anna create anna-workspace
 
 # Optional on GPU machines: create anna-workspace/.anna-deploy-venv for vLLM.
-anna init anna-workspace --deploy-env
+anna create anna-workspace --deploy-env
 
 # Fast path: use the base chat model for complaint-chain and emotion modules.
 anna models use-base --target all --workspace anna-workspace
 
 # SFT path: deploy local vLLM services from downloaded HuggingFace assets.
-anna assets pull paper --workspace anna-workspace
-anna models env setup --workspace anna-workspace  # optional if init used --deploy-env
+anna assets download paper --workspace anna-workspace
+anna models env setup --workspace anna-workspace  # optional if create used --deploy-env
 anna models deploy --target complaint --backend vllm --workspace anna-workspace \
   --gpu 0 --gpu-memory-utilization 0.85 --wait-timeout 900
 anna models deploy --target emotion --backend vllm --workspace anna-workspace \
@@ -169,7 +169,7 @@ AnnaAgent provides a Typer-based CLI organized around the reader journey from
 paper reproduction to application use. Start by creating an isolated workspace:
 
 ```bash
-anna init anna-workspace
+anna create anna-workspace
 anna doctor --workspace anna-workspace
 ```
 
@@ -188,7 +188,7 @@ anna config validate --workspace anna-workspace
 
 `config wizard` and `config secrets` use hidden password-style prompts for API
 keys and write them to `.env`. The generated `.env` and `.env.example` files
-include commented placeholders showing exactly where to put base-model, SFT and
+include commented placeholders showing exactly where to put backbone, SFT and
 embedding credentials if you prefer manual editing.
 
 Assets are manifest-driven. The default `paper` preset points to the released
@@ -197,23 +197,23 @@ HuggingFace SFT models and synthetic data, and you can override or extend it in
 
 ```bash
 anna assets list --workspace anna-workspace
-anna assets pull paper --workspace anna-workspace
+anna assets download paper --workspace anna-workspace
 ```
 
-Always pass `--workspace` or `--manifest` when pulling assets. Without either,
-`anna assets pull` uses the current directory as the workspace and may download
-to `./assets/...` instead of your intended AnnaAgent workspace. You can pull one
+Always pass `--workspace` or `--manifest` when downloading assets. Without either,
+`anna assets download` uses the current directory as the workspace and may download
+to `./assets/...` instead of your intended AnnaAgent workspace. You can download one
 specific resource or override the target directory explicitly:
 
 ```bash
-# Pull one asset from anna-workspace/assets/anna-assets.json.
-anna assets pull complaint-sft --workspace anna-workspace
+# Download one asset from anna-workspace/assets/anna-assets.json.
+anna assets download complaint-sft --workspace anna-workspace
 
 # Use an explicit manifest JSON, including absolute target paths in that file.
-anna assets pull complaint-sft --manifest anna-workspace/assets/anna-assets.json
+anna assets download complaint-sft --manifest anna-workspace/assets/anna-assets.json
 
 # Override the target directory for exactly one selected asset.
-anna assets pull complaint-sft --workspace anna-workspace \
+anna assets download complaint-sft --workspace anna-workspace \
   --target /path/to/models/complaint-sft
 ```
 
@@ -272,7 +272,7 @@ anna models deploy --target complaint --backend vllm --workspace anna-workspace
 ```
 When `--model-path` is omitted, deploy reads the corresponding SFT asset target
 from `assets/anna-assets.json`, including absolute paths. Pass the same
-`--workspace` or `--manifest` that you used during `assets pull`.
+`--workspace` or `--manifest` that you used during `assets download`.
 
 If `models deploy` reports that vLLM is unavailable, run
 `anna models env setup --workspace anna-workspace` to create the workspace deploy
@@ -299,11 +299,11 @@ Initialization can be run as a full AnnaAgent initialization, or as a prompt-onl
 state for cheap dry-runs and reproducible prompt freezing:
 
 ```bash
-anna initialize prompt-only anna-workspace/cases/family_stress_case.json \
+anna init prompt-only anna-workspace/cases/family_stress_case.json \
   --out anna-workspace/prompts/family.prompt.json
-anna initialize full anna-workspace/cases/family_stress_case.json \
+anna init full anna-workspace/cases/family_stress_case.json \
   --out anna-workspace/prompts/family.full.json --workspace anna-workspace
-anna initialize from-prompt anna-workspace/prompts/family.prompt.json
+anna init from-prompt anna-workspace/prompts/family.prompt.json
 ```
 
 Chat interactively from either a case file or a frozen prompt state:
@@ -458,8 +458,8 @@ the base model, or `true` to call the configured SFT endpoint.
 
 ```yaml
 model_service:
-  model_name: counselor
-  api_key: counselor
+  model_name: anna-backbone
+  api_key: anna-backbone
   base_url: http://localhost:8002/v1
 servers:
   complaint:

{anna_agent-0.2.0 → anna_agent-0.2.2}/README.md

@@ -37,7 +37,7 @@ After installation, the short command `anna` is available in any terminal:
 
 ```bash
 anna --version
-anna init anna-workspace
+anna create anna-workspace
 anna doctor --workspace anna-workspace
 ```
 
@@ -48,7 +48,7 @@ a workspace deployment environment on the Linux/GPU machine. This keeps the
 lightweight `anna` CLI install separate from heavy vLLM dependencies:
 
 ```bash
-anna init anna-workspace --deploy-env
+anna create anna-workspace --deploy-env
 
 # Or add it later to an existing workspace.
 anna models env setup --workspace anna-workspace
@@ -93,17 +93,17 @@ First, create a workspace, choose whether to use the SFT modules, and let the CL
 write the resulting configuration:
 
 ```bash
-anna init anna-workspace
+anna create anna-workspace
 
 # Optional on GPU machines: create anna-workspace/.anna-deploy-venv for vLLM.
-anna init anna-workspace --deploy-env
+anna create anna-workspace --deploy-env
 
 # Fast path: use the base chat model for complaint-chain and emotion modules.
 anna models use-base --target all --workspace anna-workspace
 
 # SFT path: deploy local vLLM services from downloaded HuggingFace assets.
-anna assets pull paper --workspace anna-workspace
-anna models env setup --workspace anna-workspace  # optional if init used --deploy-env
+anna assets download paper --workspace anna-workspace
+anna models env setup --workspace anna-workspace  # optional if create used --deploy-env
 anna models deploy --target complaint --backend vllm --workspace anna-workspace \
   --gpu 0 --gpu-memory-utilization 0.85 --wait-timeout 900
 anna models deploy --target emotion --backend vllm --workspace anna-workspace \
@@ -145,7 +145,7 @@ AnnaAgent provides a Typer-based CLI organized around the reader journey from
 paper reproduction to application use. Start by creating an isolated workspace:
 
 ```bash
-anna init anna-workspace
+anna create anna-workspace
 anna doctor --workspace anna-workspace
 ```
 
@@ -164,7 +164,7 @@ anna config validate --workspace anna-workspace
 
 `config wizard` and `config secrets` use hidden password-style prompts for API
 keys and write them to `.env`. The generated `.env` and `.env.example` files
-include commented placeholders showing exactly where to put base-model, SFT and
+include commented placeholders showing exactly where to put backbone, SFT and
 embedding credentials if you prefer manual editing.
 
 Assets are manifest-driven. The default `paper` preset points to the released
@@ -173,23 +173,23 @@ HuggingFace SFT models and synthetic data, and you can override or extend it in
 
 ```bash
 anna assets list --workspace anna-workspace
-anna assets pull paper --workspace anna-workspace
+anna assets download paper --workspace anna-workspace
 ```
 
-Always pass `--workspace` or `--manifest` when pulling assets. Without either,
-`anna assets pull` uses the current directory as the workspace and may download
-to `./assets/...` instead of your intended AnnaAgent workspace. You can pull one
+Always pass `--workspace` or `--manifest` when downloading assets. Without either,
+`anna assets download` uses the current directory as the workspace and may download
+to `./assets/...` instead of your intended AnnaAgent workspace. You can download one
 specific resource or override the target directory explicitly:
 
 ```bash
-# Pull one asset from anna-workspace/assets/anna-assets.json.
-anna assets pull complaint-sft --workspace anna-workspace
+# Download one asset from anna-workspace/assets/anna-assets.json.
+anna assets download complaint-sft --workspace anna-workspace
 
 # Use an explicit manifest JSON, including absolute target paths in that file.
-anna assets pull complaint-sft --manifest anna-workspace/assets/anna-assets.json
+anna assets download complaint-sft --manifest anna-workspace/assets/anna-assets.json
 
 # Override the target directory for exactly one selected asset.
-anna assets pull complaint-sft --workspace anna-workspace \
+anna assets download complaint-sft --workspace anna-workspace \
   --target /path/to/models/complaint-sft
 ```
 
@@ -248,7 +248,7 @@ anna models deploy --target complaint --backend vllm --workspace anna-workspace
 ```
 When `--model-path` is omitted, deploy reads the corresponding SFT asset target
 from `assets/anna-assets.json`, including absolute paths. Pass the same
-`--workspace` or `--manifest` that you used during `assets pull`.
+`--workspace` or `--manifest` that you used during `assets download`.
 
 If `models deploy` reports that vLLM is unavailable, run
 `anna models env setup --workspace anna-workspace` to create the workspace deploy
@@ -275,11 +275,11 @@ Initialization can be run as a full AnnaAgent initialization, or as a prompt-onl
 state for cheap dry-runs and reproducible prompt freezing:
 
 ```bash
-anna initialize prompt-only anna-workspace/cases/family_stress_case.json \
+anna init prompt-only anna-workspace/cases/family_stress_case.json \
   --out anna-workspace/prompts/family.prompt.json
-anna initialize full anna-workspace/cases/family_stress_case.json \
+anna init full anna-workspace/cases/family_stress_case.json \
   --out anna-workspace/prompts/family.full.json --workspace anna-workspace
-anna initialize from-prompt anna-workspace/prompts/family.prompt.json
+anna init from-prompt anna-workspace/prompts/family.prompt.json
 ```
 
 Chat interactively from either a case file or a frozen prompt state:
@@ -434,8 +434,8 @@ the base model, or `true` to call the configured SFT endpoint.
 
 ```yaml
 model_service:
-  model_name: counselor
-  api_key: counselor
+  model_name: anna-backbone
+  api_key: anna-backbone
   base_url: http://localhost:8002/v1
 servers:
   complaint:

{anna_agent-0.2.0 → anna_agent-0.2.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "anna-agent"
-version = "0.2.0"
+version = "0.2.2"
 description = "Dynamic Evolution Agent System with Multi-Session Memory for Realistic Seeker Simulation"
 readme = "README.md"
 requires-python = ">=3.10"

{anna_agent-0.2.0 → anna_agent-0.2.2}/src/anna_agent/__init__.py

@@ -2,4 +2,4 @@
 
 __all__ = []
 
-__version__ = "0.2.0"
+__version__ = "0.2.2"

{anna_agent-0.2.0 → anna_agent-0.2.2}/src/anna_agent/anna_agent_template.py

@@ -1,17 +1,18 @@
 ## 主诉请求和情绪通过reminder来控制
 
 prompt_template = """
-你正在扮演 心理障碍的患者，你正在cosplay 心理咨询患者。
-结合历史内容的内容用一致性的语气回复。配合我进行演出，
-请不要回答你是语言模型，永远记住你正在扮演 心理咨询患者
-注意保持你的性格特点包括 {situation}
+# Role: 心理咨询患者
 
 ## Profile
 - 性别: {gender}
 - 年龄: {age}
 - 职业: {occupation}
 - 婚姻状况: {marriage}
- 
+
+## Situation
+- 你是一个有心理障碍的患者，正在与心理咨询师进行对话。
+{situation}
+
 ## Status
 {status}
 
@@ -24,4 +25,6 @@ prompt_template = """
 ## Constraints
 - 使用中文回复
 - 一次不能提及过多的症状信息，每轮最多讨论一个症状。
+- 你应该用含糊和口语化的方式表达你的症状，并将其与你的生活经历联系起来。
+- 不要使用专业术语。
 """

{anna_agent-0.2.0 → anna_agent-0.2.2}/src/anna_agent/cli.py

@@ -79,7 +79,8 @@ app.add_typer(config_app, name="config")
 app.add_typer(test_app, name="test")
 app.add_typer(data_app, name="data")
 app.add_typer(memory_app, name="memory")
-app.add_typer(initialize_app, name="initialize")
+app.add_typer(initialize_app, name="init")
+app.add_typer(initialize_app, name="initialize", hidden=True)
 app.add_typer(models_app, name="models")
 models_app.add_typer(models_env_app, name="env")
 app.add_typer(run_app, name="run")
@@ -147,8 +148,8 @@ def _render_config_summary(
     table.add_row("Workspace", str(workspace))
     if source:
         table.add_row(source_kind, str(source))
-    table.add_row("Base model", cfg.model_name)
-    table.add_row("Base endpoint", cfg.base_url)
+    table.add_row("AnnaAgent backbone model", cfg.model_name)
+    table.add_row("Backbone endpoint", cfg.base_url)
     table.add_row("Memory", "on" if cfg.memory_enabled else "off")
     table.add_row("Debug UI", "on" if debug_ui else "off")
     console.print(Panel(table, title="配置摘要", border_style="blue"))
@@ -195,17 +196,77 @@ def _print_deploy_env_status(status: dict[str, Any]) -> None:
     console.print(table)
 
 
-def _looks_like_connection_error(err: BaseException) -> bool:
+def _iter_exception_chain(err: BaseException):
     current: BaseException | None = err
     while current is not None:
-        if current.__class__.__name__ in {"APIConnectionError", "ConnectError"}:
-            return True
+        yield current
         current = current.__cause__ or current.__context__
+
+
+def _looks_like_connection_error(err: BaseException) -> bool:
+    for current in _iter_exception_chain(err):
+        if current.__class__.__name__ in {
+            "APIConnectionError",
+            "APITimeoutError",
+            "ConnectError",
+            "ConnectTimeout",
+            "ReadTimeout",
+            "TimeoutException",
+        }:
+            return True
+    return False
+
+
+def _looks_like_auth_error(err: BaseException) -> bool:
+    for current in _iter_exception_chain(err):
+        class_name = current.__class__.__name__
+        status_code = getattr(current, "status_code", None)
+        text = str(current).lower()
+        if class_name in {"AuthenticationError", "PermissionDeniedError"}:
+            return True
+        if status_code in {401, 403}:
+            return True
+        if "invalid api key" in text or "invalid_key" in text:
+            return True
     return False
 
 
+def _looks_like_model_service_error(err: BaseException) -> bool:
+    if _looks_like_connection_error(err) or _looks_like_auth_error(err):
+        return True
+    for current in _iter_exception_chain(err):
+        if current.__class__.__name__ in {"APIStatusError", "RateLimitError"}:
+            return True
+    return False
+
+
+def _configured_secret_values() -> list[str]:
+    try:
+        cfg = registry.get("anna_engine_config")
+    except Exception:
+        return []
+    return [
+        cfg.api_key,
+        cfg.complaint_api_key,
+        cfg.counselor_api_key,
+        cfg.emotion_api_key,
+        cfg.embedding_api_key,
+    ]
+
+
 def _print_model_connection_help(workspace: Path, err: BaseException) -> None:
-    console.print(f"[red]Model service connection failed:[/red] {escape(str(err))}")
+    detail = escape(_redact_text(str(err), _configured_secret_values()))
+    if _looks_like_auth_error(err):
+        console.print(f"[red]Model service authentication failed:[/red] {detail}")
+        console.print(
+            "Check [bold]ANNA_ENGINE_API_KEY[/bold] or [bold]MIMO_API_KEY[/bold] "
+            "in the workspace .env and ensure the key matches the configured "
+            "base_url/model."
+        )
+    elif _looks_like_connection_error(err):
+        console.print(f"[red]Model service connection failed:[/red] {detail}")
+    else:
+        console.print(f"[red]Model service request failed:[/red] {detail}")
     console.print(
         "Run [bold]anna doctor --workspace {workspace}[/bold] and "
         "[bold]anna models status --workspace {workspace}[/bold]. If you use "
@@ -397,8 +458,8 @@ def doctor(
     console.print(table)
 
 
-@app.command("init")
-def init_workspace(
+@app.command("create")
+def create_workspace(
     target: Path = typer.Argument(Path("anna-workspace"), help="Workspace directory."),
     force: bool = typer.Option(False, "--force", help="Overwrite generated files."),
     deploy_env: bool = typer.Option(
@@ -418,7 +479,7 @@ def init_workspace(
     ),
 ) -> None:
     initialize_workspace(target, force=force)
-    console.print(f"[green]Workspace initialized:[/green] {target}")
+    console.print(f"[green]Workspace created:[/green] {target}")
     if deploy_env:
         try:
             status = setup_deploy_env(
@@ -464,8 +525,9 @@ def assets_list(
     console.print(table)
 
 
-@assets_app.command("pull")
-def assets_pull(
+@assets_app.command("pull", hidden=True)
+@assets_app.command("download")
+def assets_download(
     names: list[str] = typer.Argument(None, help="Asset names or preset names."),
     workspace: Path = typer.Option(Path(), "--workspace", "--root", resolve_path=True),
     force: bool = typer.Option(False, "--force", help="Redownload existing files."),
@@ -488,7 +550,7 @@ def assets_pull(
         manifest_file=manifest,
         target_override=target,
     )
-    table = Table(title="Asset Pull Results")
+    table = Table(title="Asset Download Results")
     table.add_column("Name")
     table.add_column("Status")
     table.add_column("Path")
@@ -533,11 +595,20 @@ def config_wizard(
     workspace: Path = typer.Option(Path(), "--workspace", "--root", resolve_path=True),
 ) -> None:
     data = load_settings(workspace)
+    console.print(
+        Panel(
+            "[bold]AnnaAgent backbone model[/bold]\n"
+            "This model drives AnnaAgent's internal seeker-simulation modules.",
+            border_style="cyan",
+        )
+    )
     data.setdefault("model_service", {})["model_name"] = typer.prompt(
-        "Base model name", default=data.get("model_service", {}).get("model_name", "")
+        "AnnaAgent backbone model name",
+        default=data.get("model_service", {}).get("model_name", ""),
     )
     data["model_service"]["base_url"] = typer.prompt(
-        "Base model base URL", default=data.get("model_service", {}).get("base_url", "")
+        "AnnaAgent backbone base URL",
+        default=data.get("model_service", {}).get("base_url", ""),
     )
     data.setdefault("embedding", {})["model_name"] = typer.prompt(
         "Embedding model", default=data.get("embedding", {}).get("model_name", "")
@@ -563,14 +634,14 @@ def config_secrets(
 
 def _prompt_and_write_secrets(workspace: Path, include_sft: bool = False) -> None:
     prompts = [
-        ("ANNA_ENGINE_API_KEY", "Base model API key"),
+        ("ANNA_ENGINE_API_KEY", "AnnaAgent backbone API key"),
         ("ANNA_ENGINE_EMBEDDING_API_KEY", "Embedding API key"),
     ]
     if include_sft:
         prompts.extend(
             [
                 ("ANNA_ENGINE_COMPLAINT_API_KEY", "Complaint SFT API key"),
-                ("ANNA_ENGINE_COUNSELOR_API_KEY", "Counselor API key"),
+                ("ANNA_ENGINE_COUNSELOR_API_KEY", "Backbone fallback API key"),
                 ("ANNA_ENGINE_EMOTION_API_KEY", "Emotion SFT API key"),
             ]
         )
@@ -612,10 +683,16 @@ def test_model(
     _configure(workspace)
     cfg = registry.get("anna_engine_config")
     client = backbone.get_openai_client()
-    response = client.chat.completions.create(
-        model=cfg.model_name,
-        messages=[{"role": "user", "content": "Reply with OK."}],
-    )
+    try:
+        response = client.chat.completions.create(
+            model=cfg.model_name,
+            messages=[{"role": "user", "content": "Reply with OK."}],
+        )
+    except Exception as err:
+        if _looks_like_model_service_error(err):
+            _print_model_connection_help(workspace, err)
+            raise typer.Exit(code=1) from None
+        raise
     console.print(
         f"[green]Model OK[/green] response={response.choices[0].message.content}"
     )
@@ -796,7 +873,10 @@ def initialize_prompt_only(
     case_file: Path = typer.Argument(..., help="Case file."),
     output: Path = typer.Option(Path("prompts/prompt_state.json"), "--out", "-o"),
 ) -> None:
+    console.print(Rule("[bold blue]Initialization · prompt-only[/bold blue]"))
+    console.print(f"[blue]Running:[/blue] load case and build prompt for {case_file}")
     state = build_prompt_only_state(case_file)
+    console.print(f"[blue]Running:[/blue] write initialization state to {output}")
     save_state(state, output)
     console.print(f"[green]Wrote prompt-only state[/green] {output}")
 
@@ -808,13 +888,26 @@ def initialize_full(
     workspace: Path = typer.Option(Path(), "--workspace", "--root", resolve_path=True),
 ) -> None:
     _configure(workspace)
+    events: list[tuple[str, str]] = []
+
+    def progress(stage: str, detail: str) -> None:
+        events.append((stage, detail))
+        console.print(f"[blue]Running {escape(stage)}:[/blue] {escape(detail)}")
+
+    console.print(Rule("[bold blue]Initialization · full[/bold blue]"))
+    console.print(
+        f"[blue]Running:[/blue] load case and configure services for {case_file}"
+    )
     try:
-        state = build_full_state(case_file)
+        state = build_full_state(case_file, progress_callback=progress)
     except Exception as err:
-        if _looks_like_connection_error(err):
+        if _looks_like_model_service_error(err):
             _print_model_connection_help(workspace, err)
             raise typer.Exit(code=1) from None
         raise
+    if events:
+        _render_initialization_events(events)
+    console.print(f"[blue]Running:[/blue] write initialization state to {output}")
     save_state(state, output)
     console.print(f"[green]Wrote full state[/green] {output}")
 
@@ -967,7 +1060,7 @@ def models_deploy(
         None, help="Override vLLM max model length."
     ),
     pull: bool = typer.Option(
-        True, "--pull/--no-pull", help="Download default asset if missing."
+        True, "--download/--no-download", help="Download default asset if missing."
     ),
     background: bool = typer.Option(True, "--background/--foreground"),
     dry_run: bool = typer.Option(
@@ -1121,11 +1214,17 @@ def chat(
     console.print(Rule("[bold blue]Stage 1/2 · Initialize[/bold blue]"))
     _configure(workspace)
     init_events: list[tuple[str, str]] = []
+    compact_status: Any = None
 
     def progress(stage: str, detail: str) -> None:
         init_events.append((stage, detail))
         if debug_ui:
             console.print(f"[dim]• {escape(stage)}[/dim] {escape(detail)}")
+        else:
+            message = f"正在运行：{detail}"
+            if compact_status is not None:
+                compact_status.update(f"[blue]{escape(message)}...[/blue]")
+            console.print(f"[blue]{escape(message)}[/blue]")
 
     if state:
         _render_config_summary(
@@ -1151,22 +1250,26 @@ def chat(
                     progress_callback=progress,
                 )
             except Exception as err:
-                if _looks_like_connection_error(err):
+                if _looks_like_model_service_error(err):
                     _print_model_connection_help(workspace, err)
                     raise typer.Exit(code=1) from None
                 raise
         else:
             try:
-                with console.status("[blue]初始化来访者状态...[/blue]"):
-                    with _suppress_stdio(True):
-                        seeker = MsPatient(
-                            portrait,
-                            report,
-                            conversations,
-                            progress_callback=progress,
-                        )
+                with console.status(
+                    "[blue]正在运行：初始化来访者状态...[/blue]"
+                ) as status:
+                    compact_status = status
+                    seeker = MsPatient(
+                        portrait,
+                        report,
+                        conversations,
+                        progress_callback=progress,
+                    )
+                    compact_status = None
             except Exception as err:
-                if _looks_like_connection_error(err):
+                compact_status = None
+                if _looks_like_model_service_error(err):
                     _print_model_connection_help(workspace, err)
                     raise typer.Exit(code=1) from None
                 raise
@@ -1215,12 +1318,18 @@ def run_batch(
     output.mkdir(parents=True, exist_ok=True)
     summary_path = output / "summary.jsonl"
     for case_file in case_files:
-        state = (
-            build_full_state(case_file)
-            if mode == "full"
-            else build_prompt_only_state(case_file)
-        )
+        console.print(f"[blue]Running:[/blue] build {mode} state for {case_file}")
+        if mode == "full":
+            state = build_full_state(
+                case_file,
+                progress_callback=lambda stage, detail: console.print(
+                    f"[blue]Running {escape(stage)}:[/blue] {escape(detail)}"
+                ),
+            )
+        else:
+            state = build_prompt_only_state(case_file)
         state_path = output / f"{state['case_id']}.state.json"
+        console.print(f"[blue]Running:[/blue] write state to {state_path}")
         save_state(state, state_path)
         record = {
             "case_id": state["case_id"],
@@ -1230,7 +1339,11 @@ def run_batch(
         if live and messages:
             session = FrozenPromptSession(state)
             transcript_path = output / f"{state['case_id']}.transcript.jsonl"
-            for message in messages:
+            for index, message in enumerate(messages, start=1):
+                console.print(
+                    f"[blue]Running:[/blue] live scripted turn {index} "
+                    f"for {state['case_id']}"
+                )
                 response = session.chat(message)
                 append_jsonl(transcript_path, {"role": "Counselor", "content": message})
                 append_jsonl(transcript_path, {"role": "Seeker", "content": response})