thwip-cli 1.4.0__tar.gz → 1.5.0__tar.gz

{thwip_cli-1.4.0 → thwip_cli-1.5.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: thwip-cli
-Version: 1.4.0
+Version: 1.5.0
 Summary: Universal coding agent multiplexer: detect, switch, and route between AI coding agents seamlessly
 Project-URL: Homepage, https://github.com/tanmayhutt/thwip-cli
 Project-URL: Repository, https://github.com/tanmayhutt/thwip-cli
@@ -54,7 +54,8 @@ Description-Content-Type: text/markdown
 - **Rate Limit Failover**: Detects HTTP 429 errors or quota exhaustion and prompts instant switching to ready fallback models.
 - **Tool Layer**: Shared file editing, code execution (Python, Node, Shell), and Git integration across connected models.
 - **Session Persistence**: Save and resume sessions across projects with `/session save` and `/session load`.
-- **Terminal Ergonomics**: Autocompletion (Tab), keyboard shortcuts (Ctrl+S, Ctrl+T, Ctrl+H), and streaming markdown rendering.
+- **Familiar Flow**: The same everyday commands as Codex, Claude Code, and Antigravity (`/model`, `/new`, `/resume`, `/compact`, `/diff`, `/copy`, `!cmd`, `@file`), plus autocompletion, Ctrl+S / Ctrl+T shortcuts, and streaming Markdown.
+- **Live Model Lists**: Model catalogs come from the connected CLI or the provider's list-models endpoint, never from a hardcoded list.
 
 ---
 
@@ -99,8 +100,17 @@ thwip --version
 | `/models [agent]` | List available models for current or target agent |
 | `/key [provider]` | Enter an API key securely without placing it in prompt history |
 | `/status` | Display current session, project, and token stats |
-| `/limits` | View token usage, quota, and spend metrics |
+| `/limits`, `/usage` | View token usage, CLI account usage windows, and spend metrics |
 | `/detect` | Re-scan installed agents and reconnect CLI sign-ins |
+| `/model [id]` | Pick a model for the current agent from a numbered list or by ID |
+| `/new` | Start a fresh conversation; the current one is saved first |
+| `/resume [name]` | Resume a saved session from a numbered list |
+| `/compact` | Summarize the conversation with the current model to free context; the summary is plain text and travels across providers |
+| `/diff [staged]` | Show the project's git diff |
+| `/copy` | Copy the last response to the clipboard |
+| `/export [path]` | Write the conversation to Markdown with per-message model attribution |
+| `!<command>` | Run a shell command in the project without involving a model |
+| `@path` in a message | Attach a project file's content to the message (files outside the project are ignored) |
 | `/session save [name]` | Save current chat session |
 | `/session load <name>` | Load a previously saved session |
 | `/session list` | List all saved sessions |
@@ -154,6 +164,28 @@ replaces Thwip with the Codex CLI itself in the selected project. Conversation
 history is not transferred by the launcher; use `/session load` after restarting
 Thwip to resume.
 
+## Live model lists
+
+Model lists are not hardcoded. Each connected source supplies its own list:
+
+- Installed CLIs report their models (Codex `model/list`, `agy models`, Claude Code aliases).
+- Direct providers with a key are queried through their list-models endpoint (OpenAI,
+  Anthropic, Google, DeepSeek, Groq, OpenRouter) at startup, on `/models`, and right
+  after `/key`. Non-chat models (embeddings, speech, image, moderation) are filtered
+  out. Context sizes and prices are taken from the provider when it publishes them.
+- A small bundled list per provider remains only as an offline fallback and is labelled
+  "bundled fallback" in `/models` until a key or connection is available.
+
+## Usage-limit failover
+
+When the active provider reports an exhausted limit (Codex "You've hit your usage
+limit", Claude Code "usage limit reached" or a rejected rate-limit event, Google
+`RESOURCE_EXHAUSTED`, or HTTP 429 from a direct API), Thwip stops, shows the ready
+alternatives with their default models, and offers to switch. Accepting switches the
+provider, keeps the text conversation, and re-sends the unanswered message. Set
+`[limits] auto_switch = true` to skip the question. The `[fallback] chain` order is
+honored; a chain model the provider does not list falls back to that provider's default.
+
 ## Auditable handoffs
 
 ```text

{thwip_cli-1.4.0 → thwip_cli-1.5.0}/README.md

@@ -17,7 +17,8 @@
 - **Rate Limit Failover**: Detects HTTP 429 errors or quota exhaustion and prompts instant switching to ready fallback models.
 - **Tool Layer**: Shared file editing, code execution (Python, Node, Shell), and Git integration across connected models.
 - **Session Persistence**: Save and resume sessions across projects with `/session save` and `/session load`.
-- **Terminal Ergonomics**: Autocompletion (Tab), keyboard shortcuts (Ctrl+S, Ctrl+T, Ctrl+H), and streaming markdown rendering.
+- **Familiar Flow**: The same everyday commands as Codex, Claude Code, and Antigravity (`/model`, `/new`, `/resume`, `/compact`, `/diff`, `/copy`, `!cmd`, `@file`), plus autocompletion, Ctrl+S / Ctrl+T shortcuts, and streaming Markdown.
+- **Live Model Lists**: Model catalogs come from the connected CLI or the provider's list-models endpoint, never from a hardcoded list.
 
 ---
 
@@ -62,8 +63,17 @@ thwip --version
 | `/models [agent]` | List available models for current or target agent |
 | `/key [provider]` | Enter an API key securely without placing it in prompt history |
 | `/status` | Display current session, project, and token stats |
-| `/limits` | View token usage, quota, and spend metrics |
+| `/limits`, `/usage` | View token usage, CLI account usage windows, and spend metrics |
 | `/detect` | Re-scan installed agents and reconnect CLI sign-ins |
+| `/model [id]` | Pick a model for the current agent from a numbered list or by ID |
+| `/new` | Start a fresh conversation; the current one is saved first |
+| `/resume [name]` | Resume a saved session from a numbered list |
+| `/compact` | Summarize the conversation with the current model to free context; the summary is plain text and travels across providers |
+| `/diff [staged]` | Show the project's git diff |
+| `/copy` | Copy the last response to the clipboard |
+| `/export [path]` | Write the conversation to Markdown with per-message model attribution |
+| `!<command>` | Run a shell command in the project without involving a model |
+| `@path` in a message | Attach a project file's content to the message (files outside the project are ignored) |
 | `/session save [name]` | Save current chat session |
 | `/session load <name>` | Load a previously saved session |
 | `/session list` | List all saved sessions |
@@ -117,6 +127,28 @@ replaces Thwip with the Codex CLI itself in the selected project. Conversation
 history is not transferred by the launcher; use `/session load` after restarting
 Thwip to resume.
 
+## Live model lists
+
+Model lists are not hardcoded. Each connected source supplies its own list:
+
+- Installed CLIs report their models (Codex `model/list`, `agy models`, Claude Code aliases).
+- Direct providers with a key are queried through their list-models endpoint (OpenAI,
+  Anthropic, Google, DeepSeek, Groq, OpenRouter) at startup, on `/models`, and right
+  after `/key`. Non-chat models (embeddings, speech, image, moderation) are filtered
+  out. Context sizes and prices are taken from the provider when it publishes them.
+- A small bundled list per provider remains only as an offline fallback and is labelled
+  "bundled fallback" in `/models` until a key or connection is available.
+
+## Usage-limit failover
+
+When the active provider reports an exhausted limit (Codex "You've hit your usage
+limit", Claude Code "usage limit reached" or a rejected rate-limit event, Google
+`RESOURCE_EXHAUSTED`, or HTTP 429 from a direct API), Thwip stops, shows the ready
+alternatives with their default models, and offers to switch. Accepting switches the
+provider, keeps the text conversation, and re-sends the unanswered message. Set
+`[limits] auto_switch = true` to skip the question. The `[fallback] chain` order is
+honored; a chain model the provider does not list falls back to that provider's default.
+
 ## Auditable handoffs
 
 ```text

{thwip_cli-1.4.0 → thwip_cli-1.5.0}/docs/verification.md

@@ -1,7 +1,7 @@
 # Verification status
 
-Verified locally on 2026-09-24 for the unreleased v1.4.0 source. The Python suite
-passes 216 offline tests. Exhaustive behavior across every provider and
+Verified locally on 2026-09-24 for v1.5.0. The Python suite
+passes 240 offline tests. Exhaustive behavior across every provider and
 configuration has not been established.
 
 ## Native CLI connections (2026-09-24)
@@ -17,6 +17,10 @@ each using its existing sign-in. No API keys were configured.
 | Context across `/switch` | Codex and Antigravity both recalled the answer given by the previous provider |
 | Codex approval request | A write command outside the read-only sandbox produced a permission prompt; denial left the workspace unchanged |
 | Ctrl+C during a response | Turn cancelled, child process terminated, REPL continued, unanswered message removed |
+| Ctrl+C at a Codex permission prompt | Turn cancelled, no file created, no leftover process, REPL continued |
+| Usage-limit failover | With a test-only shim making Codex report "You've hit your usage limit", the real REPL showed the alternatives, switched to Claude Code on `1`, retried the message, and answered; history held one clean pair |
+| Parity commands | Live REPL run: `!git log`, `/model` picker, `@file` mention answered by Codex, `/compact` summary, `/export`, `/copy`, `/diff`, `/new`, `/resume`, `/usage` |
+| Live catalogs | OpenRouter public list fetched live (459 models with context and pricing); other providers covered by recorded-payload tests because no keys are present here |
 | `/session save` and `/session load` | Session with a native provider saved and reloaded in a fresh run |
 | `/models`, `/models <provider>`, `/models <tier>` | Live catalogs listed with `CLI account` in place of API pricing |
 | `thwip --version`, `--help`, `--project` | Handled without starting the REPL; invalid project exits with code 2 |
@@ -30,7 +34,7 @@ error. The protocol enum is `read-only`. A regression test now checks the reques
 Remaining limitations: the Antigravity CLI showed intermittent network resets to
 Google's backend during testing, which surface as turn errors; Claude Code's model
 aliases are a curated list because the CLI exposes no model listing; native usage
-limit failover is unit-tested from error text, not observed live; the real Gemini
+limit failover was observed live only through an injected limit, not a real provider cap; the real Gemini
 CLI ACP path is covered by mocked tests only because it is not installed here.
 
 | Area | Evidence | Remaining limitation |

{thwip_cli-1.4.0 → thwip_cli-1.5.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "thwip-cli"
-version = "1.4.0"
+version = "1.5.0"
 description = "Universal coding agent multiplexer: detect, switch, and route between AI coding agents seamlessly"
 readme = "README.md"
 license = "MIT"

{thwip_cli-1.4.0 → thwip_cli-1.5.0}/tests/test_audit_regressions.py

@@ -6,7 +6,7 @@ import pytest
 from rich.console import Console
 
 from thwip.agents import AgentRegistry
-from thwip.agents.base import AgentDone, Capability, LimitHit, LimitStatus, TextDelta
+from thwip.agents.base import AgentDone, Capability, LimitHit, LimitStatus, ModelInfo, TextDelta
 from thwip.cli import ThwipCLI
 from thwip.config import ThwipConfig, get_usage_path
 from thwip.limits import UsageTracker
@@ -203,3 +203,62 @@ async def test_failed_turn_removes_unanswered_user_message(cli, monkeypatch):
     cli.session.current_agent = 'openai'
     await cli.process_user_message('hello')
     assert cli.session.messages == []
+
+
+@pytest.mark.asyncio
+async def test_ctrl_c_at_native_permission_prompt_cancels_turn_and_closes_adapter(cli, monkeypatch):
+    from thwip.agents.base import NativePermission
+    from thwip.cli import TurnInterrupted
+
+    closed = []
+
+    class NativeAsking:
+        name = 'openai'
+        display_name = 'Native'
+        company = 'OpenAI'
+        native_tools = True
+        project = '.'
+        def is_configured(self):
+            return True
+        def is_installed(self):
+            return True
+        def get_capabilities_for_model(self, model):
+            return set()
+        async def chat(self, **kwargs):
+            try:
+                yield NativePermission('run something')
+                yield TextDelta(content='never reached')
+            finally:
+                closed.append(True)
+
+    async def interrupted(self, question):
+        raise TurnInterrupted
+
+    monkeypatch.setattr(ThwipCLI, '_ask_yes_no', interrupted)
+    cli.current_agent = NativeAsking()
+    cli.session.current_agent = 'openai'
+    await cli._run_interruptible(cli.process_user_message('hello'))
+    assert cli.session.messages == []
+    assert closed == [True]
+
+
+@pytest.mark.asyncio
+async def test_failover_chain_model_must_be_listed_by_provider(cli, monkeypatch):
+    """A chain entry like claude/claude-opus-5 must not force an unlisted model onto a native adapter."""
+    from thwip.agents.native_print import PrintAgent
+
+    native = PrintAgent('claude', '.')
+    native.ready = True
+    native.available_models = [ModelInfo(id='fable', name='Fable', is_default=True)]
+    monkeypatch.setattr(native, 'is_installed', lambda: True)
+    cli.registry._agents['claude'] = native
+    cli.config.fallback.chain = ['claude/claude-opus-5']
+    switched = []
+
+    async def fake_switch(name, model=''):
+        switched.append((name, model))
+    monkeypatch.setattr(cli, 'cmd_switch', fake_switch)
+    monkeypatch.setattr('builtins.input', lambda *args: '1')
+    cli.current_agent = cli.registry.get_agent('openai')
+    await cli.handle_limit_failover(LimitHit(error_type=LimitStatus.QUOTA_EXHAUSTED, message='usage limit'), {'openai'})
+    assert switched == [('claude', 'fable')]

thwip_cli-1.5.0/tests/test_catalog.py

@@ -0,0 +1,131 @@
+"""Live model catalogs, exercised against recorded provider payload shapes only."""
+
+
+import httpx
+import pytest
+
+from thwip.agents import AgentRegistry
+from thwip.agents.base import ModelInfo
+from thwip.agents.catalog import fetch_live_models, merge_catalog, parse_models, supports_live_catalog, tier_for
+from thwip.agents.openai_agent import OpenAIAgent
+from thwip.config import ThwipConfig
+
+PAYLOADS = {
+    "openai": {"data": [{"id": "gpt-5.6-sol"}, {"id": "gpt-5.7-astra"}, {"id": "text-embedding-3-large"},
+                         {"id": "whisper-1"}, {"id": "gpt-4o-realtime-preview"}, {"id": "o4-mini"}]},
+    "claude": {"data": [{"id": "claude-fable-5-1", "display_name": "Claude Fable 5.1"}, {"id": "claude-haiku-4-5", "display_name": "Claude Haiku 4.5"}]},
+    "google": {"models": [
+        {"name": "models/gemini-3.7-flash", "displayName": "Gemini 3.7 Flash", "supportedGenerationMethods": ["generateContent"],
+         "inputTokenLimit": 1048576, "outputTokenLimit": 65536},
+        {"name": "models/text-embedding-004", "supportedGenerationMethods": ["embedContent"]},
+        {"name": "models/imagen-4", "supportedGenerationMethods": ["generateContent"]}]},
+    "deepseek": {"data": [{"id": "deepseek-chat"}, {"id": "deepseek-reasoner"}]},
+    "groq": {"data": [{"id": "openai/gpt-oss-120b", "context_window": 131072}, {"id": "whisper-large-v3"}, {"id": "llama-guard-4"}]},
+    "openrouter": {"data": [{"id": "anthropic/claude-opus-5", "name": "Anthropic: Claude Opus 5", "context_length": 200000,
+                             "pricing": {"prompt": "0.000015", "completion": "0.000075"}, "top_provider": {"max_completion_tokens": 32000}}]},
+}
+
+
+@pytest.mark.parametrize("provider,expected", [
+    ("openai", ["gpt-5.6-sol", "gpt-5.7-astra", "o4-mini"]),
+    ("claude", ["claude-fable-5-1", "claude-haiku-4-5"]),
+    ("google", ["gemini-3.7-flash"]),
+    ("deepseek", ["deepseek-chat", "deepseek-reasoner"]),
+    ("groq", ["openai/gpt-oss-120b"]),
+    ("openrouter", ["anthropic/claude-opus-5"]),
+])
+def test_parse_keeps_only_chat_models(provider, expected):
+    assert [m.id for m in parse_models(provider, PAYLOADS[provider])] == expected
+
+
+def test_parse_carries_context_and_pricing():
+    google = parse_models("google", PAYLOADS["google"])[0]
+    assert (google.context_window, google.max_output, google.name) == (1048576, 65536, "Gemini 3.7 Flash")
+    router = parse_models("openrouter", PAYLOADS["openrouter"])[0]
+    assert (router.pricing_input, router.pricing_output, router.max_output) == (15.0, 75.0, 32000)
+    assert tier_for("gemini-3.1-pro-high") == "flagship" and tier_for("gpt-5-mini") == "fast"
+
+
+@pytest.mark.asyncio
+async def test_fetch_sends_provider_auth_and_rejects_bad_keys():
+    seen = {}
+
+    def handler(request):
+        seen["url"] = str(request.url)
+        seen["headers"] = dict(request.headers)
+        if "anthropic" in seen["url"]:
+            return httpx.Response(401, json={"error": "bad key"})
+        return httpx.Response(200, json=PAYLOADS["openai"])
+
+    async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as client:
+        models = await fetch_live_models("openai", "sk-test", client)
+        assert [m.id for m in models][:2] == ["gpt-5.6-sol", "gpt-5.7-astra"]
+        assert seen["headers"]["authorization"] == "Bearer sk-test"
+        with pytest.raises(PermissionError, match="rejected"):
+            await fetch_live_models("claude", "bad", client)
+        assert seen["headers"]["x-api-key"] == "bad" and "anthropic-version" in seen["headers"]
+    assert supports_live_catalog("openai") and not supports_live_catalog("ollama")
+
+
+def test_merge_uses_live_list_as_truth_and_keeps_known_metadata():
+    bundled = [ModelInfo(id="gpt-5.6-sol", name="GPT-5.6 Sol", tier="balanced", pricing_input=1.25, pricing_output=10.0, is_default=True),
+               ModelInfo(id="gpt-retired", name="Retired")]
+    live = [ModelInfo(id="gpt-5.7-astra", name="gpt-5.7-astra"), ModelInfo(id="gpt-5.6-sol", name="gpt-5.6-sol", context_window=400000)]
+    merged = merge_catalog(bundled, live)
+    assert [m.id for m in merged] == ["gpt-5.6-sol", "gpt-5.7-astra"]
+    sol = merged[0]
+    assert sol.is_default and sol.pricing_input == 1.25 and sol.context_window == 400000 and sol.name == "GPT-5.6 Sol"
+    assert not merged[1].is_default
+    # When the bundled default disappears, the first listed model becomes the default.
+    assert merge_catalog(bundled, [ModelInfo(id="gpt-new", name="new")])[0].is_default
+
+
+@pytest.mark.asyncio
+async def test_agent_refresh_replaces_bundled_catalog_when_key_is_set(monkeypatch):
+    async def fake_fetch(provider, key, client=None):
+        assert (provider, key) == ("openai", "sk-live")
+        return parse_models("openai", PAYLOADS["openai"])
+
+    monkeypatch.setattr("thwip.agents.catalog.fetch_live_models", fake_fetch)
+    agent = OpenAIAgent(api_key="sk-live")
+    assert agent.catalog_source == "bundled"
+    await agent.refresh_models()
+    assert agent.catalog_source == "live" and agent.get_model_info("gpt-5.7-astra") is not None
+    assert agent.get_model_info("gpt-5.6-luna") is None, "models the provider no longer lists are dropped"
+    assert OpenAIAgent.available_models[0].id == "gpt-5.6-sol", "class-level bundled fallback is untouched"
+
+    unkeyed = OpenAIAgent(api_key=None)
+    monkeypatch.delenv("OPENAI_API_KEY", raising=False)
+    await unkeyed.refresh_models()
+    assert unkeyed.catalog_source == "bundled"
+
+
+@pytest.mark.asyncio
+async def test_agent_refresh_failure_keeps_fallback_with_message(monkeypatch):
+    async def failing(provider, key, client=None):
+        raise httpx.ConnectError("boom sk-" + "x" * 40)
+
+    monkeypatch.setattr("thwip.agents.catalog.fetch_live_models", failing)
+    agent = OpenAIAgent(api_key="sk-live")
+    await agent.refresh_models()
+    assert agent.catalog_source == "bundled" and "bundled fallback" in agent.discovery_error
+    assert "xxxx" not in agent.discovery_error and agent.available_models
+
+
+@pytest.mark.asyncio
+async def test_registry_refreshes_keyed_direct_adapters_at_connect(monkeypatch, tmp_path):
+    config = ThwipConfig(project=str(tmp_path))
+    config.keys = {"deepseek": "sk-deep"}
+    registry = AgentRegistry(config)
+    monkeypatch.setattr("thwip.agents.native_agent.shutil.which", lambda name: None)
+    monkeypatch.setattr("thwip.agents.native_print.shutil.which", lambda name: None)
+    calls = []
+
+    async def fake_fetch(provider, key, client=None):
+        calls.append(provider)
+        return parse_models(provider, PAYLOADS[provider])
+
+    monkeypatch.setattr("thwip.agents.catalog.fetch_live_models", fake_fetch)
+    await registry.connect_native_agents(str(tmp_path))
+    assert calls == ["deepseek"]
+    assert [m.id for m in registry.get_agent("deepseek").available_models] == ["deepseek-chat", "deepseek-reasoner"]

thwip_cli-1.5.0/tests/test_parity_commands.py

@@ -0,0 +1,173 @@
+"""Commands that mirror Codex, Claude Code, and Antigravity REPL flows, tested offline."""
+
+import subprocess
+
+import pytest
+
+from thwip.agents import AgentRegistry
+from thwip.agents.base import AgentDone, LimitHit, LimitStatus, ModelInfo, TextDelta
+from thwip.cli import ThwipCLI
+from thwip.config import ThwipConfig
+from thwip.limits import UsageTracker
+from thwip.session import Session
+from thwip.tools import ToolManager
+
+
+class FakeAgent:
+    name = "openai"
+    display_name = "Fake"
+    company = "OpenAI"
+    native_tools = False
+    capabilities = set()
+
+    def __init__(self, reply="- Context: built a CLI\n- Decisions: use pty\n- Open tasks: release"):
+        self.available_models = [ModelInfo(id="alpha", name="Alpha", is_default=True), ModelInfo(id="beta", name="Beta", tier="fast")]
+        self.reply = reply
+        self.requests = []
+
+    def is_configured(self):
+        return True
+
+    def is_installed(self):
+        return True
+
+    def get_model_info(self, model_id):
+        return next((m for m in self.available_models if m.id == model_id), None)
+
+    def get_default_model(self):
+        return "alpha"
+
+    def get_capabilities_for_model(self, model):
+        return set()
+
+    async def chat(self, **kwargs):
+        self.requests.append(kwargs)
+        if self.reply == "LIMIT":
+            yield LimitHit(error_type=LimitStatus.QUOTA_EXHAUSTED, message="usage limit")
+            return
+        yield TextDelta(content=self.reply)
+        yield AgentDone()
+
+
+@pytest.fixture
+def cli(tmp_path, monkeypatch):
+    monkeypatch.setenv("THWIP_CONFIG_DIR", str(tmp_path / "config"))
+    cli = ThwipCLI.__new__(ThwipCLI)
+    cli.config = ThwipConfig(project=str(tmp_path), auto_save=True)
+    cli.registry = AgentRegistry(cli.config)
+    cli.usage_tracker = UsageTracker()
+    cli.tool_manager = ToolManager(str(tmp_path))
+    cli.session = Session(project_path=str(tmp_path), current_agent="openai", current_model="alpha")
+    cli.current_agent = FakeAgent()
+    return cli
+
+
+def answers(monkeypatch, cli, *values):
+    queue = list(values)
+
+    async def ask(question):
+        return queue.pop(0)
+    monkeypatch.setattr(cli, "_ask_text", ask)
+
+
+@pytest.mark.asyncio
+async def test_model_picker_by_number_id_and_unknown(cli, monkeypatch):
+    answers(monkeypatch, cli, "2")
+    await cli.cmd_model()
+    assert cli.session.current_model == "beta"
+    await cli.cmd_model("alpha")
+    assert cli.session.current_model == "alpha"
+    await cli.cmd_model("nope")
+    assert cli.session.current_model == "alpha"
+    answers(monkeypatch, cli, "")
+    await cli.cmd_model()
+    assert cli.session.current_model == "alpha"
+
+
+@pytest.mark.asyncio
+async def test_new_saves_previous_and_resume_restores_it(cli, monkeypatch):
+    cli.session.add_user_message("hello")
+    cli.session.add_assistant_message("hi", agent_name="openai", model="alpha")
+    old_name = cli.session.name
+    cli.cmd_new()
+    assert cli.session.messages == [] and cli.session.name != old_name
+    assert Session.load(old_name) is not None
+    answers(monkeypatch, cli, "1")
+    await cli.cmd_resume()
+    assert cli.session.name == old_name and len(cli.session.messages) == 2
+    cli.cmd_new()
+    await cli.cmd_resume(old_name)
+    assert cli.session.name == old_name
+
+
+@pytest.mark.asyncio
+async def test_compact_replaces_history_with_portable_summary(cli):
+    cli.session.add_user_message("build a cli")
+    cli.session.add_assistant_message("done with pty", agent_name="openai", model="alpha")
+    cli.session.add_user_message("now release")
+    cli.session.add_assistant_message("ok", agent_name="openai", model="alpha")
+    await cli.cmd_compact()
+    assert len(cli.session.messages) == 2
+    assert "Decisions: use pty" in cli.session.messages[0].content and cli.session.messages[0].role == "user"
+    assert cli.session.messages[1].role == "assistant" and cli.session.messages[1].model == "alpha"
+    sent = cli.current_agent.requests[0]
+    assert "now release" in sent["messages"][0]["content"] and sent["tools"] is None
+
+
+@pytest.mark.asyncio
+async def test_compact_keeps_history_on_limit_or_empty(cli):
+    cli.session.add_user_message("a")
+    cli.session.add_assistant_message("b", agent_name="openai", model="alpha")
+    cli.current_agent = FakeAgent(reply="LIMIT")
+    await cli.cmd_compact()
+    assert len(cli.session.messages) == 2
+    cli.current_agent = FakeAgent(reply="   ")
+    await cli.cmd_compact()
+    assert len(cli.session.messages) == 2
+
+
+def test_diff_uses_project_repo(cli, tmp_path):
+    subprocess.run(["git", "init", "-q"], cwd=tmp_path, check=True)
+    (tmp_path / "a.txt").write_text("one\n")
+    subprocess.run(["git", "add", "a.txt"], cwd=tmp_path, check=True)
+    cli.cmd_diff("")          # clean working tree message
+    (tmp_path / "a.txt").write_text("two\n")
+    cli.cmd_diff("")          # renders a diff
+    cli.cmd_diff("staged")    # staged view
+
+
+def test_copy_uses_available_clipboard_tool(cli, monkeypatch):
+    calls = []
+    monkeypatch.setattr("thwip.cli.shutil.which", lambda name: "/usr/bin/pbcopy" if name == "pbcopy" else None)
+    monkeypatch.setattr(subprocess, "run", lambda command, **kwargs: calls.append((command, kwargs["input"])))
+    cli.cmd_copy()
+    assert calls == []
+    cli.session.add_assistant_message("copy me", agent_name="openai", model="alpha")
+    cli.cmd_copy()
+    assert calls == [(["pbcopy"], b"copy me")]
+
+
+def test_export_writes_markdown_with_attribution(cli, tmp_path):
+    cli.session.add_user_message("question")
+    cli.session.add_assistant_message("answer", agent_name="openai", model="alpha", company="OpenAI")
+    cli.cmd_export("notes/out.md")
+    text = (tmp_path / "notes" / "out.md").read_text()
+    assert "## You\n\nquestion" in text and "## Assistant (OpenAI / alpha)\n\nanswer" in text
+    cli.cmd_export("")
+    assert (tmp_path / f"{cli.session.name}.md").exists()
+
+
+@pytest.mark.asyncio
+async def test_shell_escape_runs_in_project(cli, tmp_path, capsys):
+    (tmp_path / "marker.txt").write_text("x")
+    await cli.cmd_shell("ls")
+    assert cli.session.messages == []
+    await cli.cmd_shell("")
+
+
+def test_mentions_attach_project_files_only(cli, tmp_path):
+    (tmp_path / "notes.md").write_text("secret plan")
+    expanded = cli._expand_mentions("Review @notes.md and @missing.txt and @../etc/passwd please")
+    assert "secret plan" in expanded and "[Attached file: notes.md]" in expanded
+    assert "missing.txt]" not in expanded and "passwd]" not in expanded
+    assert cli._expand_mentions("email me at user@example.com") == "email me at user@example.com"

{thwip_cli-1.4.0 → thwip_cli-1.5.0}/thwip/__init__.py

@@ -1,4 +1,4 @@
 """thwip: Universal Coding Agent Multiplexer."""
 
-__version__ = "1.4.0"
+__version__ = "1.5.0"
 __app_name__ = "thwip"

{thwip_cli-1.4.0 → thwip_cli-1.5.0}/thwip/agents/__init__.py

@@ -99,11 +99,16 @@ class AgentRegistry:
                 self._agents[name] = native
             native.project = project
             natives.append(native)
-        results = await asyncio.gather(*(native.refresh_models() for native in natives), return_exceptions=True)
-        for native, result in zip(natives, results):
+        # Direct adapters with a key fetch their live model list at the same time.
+        direct = [agent for agent in self._agents.values()
+                  if not getattr(agent, "native_tools", False) and agent.name != "ollama" and agent.is_configured()]
+        refreshing = natives + direct
+        results = await asyncio.gather(*(agent.refresh_models() for agent in refreshing), return_exceptions=True)
+        for agent, result in zip(refreshing, results):
             if isinstance(result, BaseException):
-                native.ready = False
-                native.discovery_error = f"Native discovery failed: {type(result).__name__}"
+                if getattr(agent, "native_tools", False):
+                    agent.ready = False
+                agent.discovery_error = f"Model discovery failed: {type(result).__name__}"
 
     @staticmethod
     def _native_candidate(name: str, project: str) -> BaseAgent | None:

{thwip_cli-1.4.0 → thwip_cli-1.5.0}/thwip/agents/base.py

@@ -316,6 +316,34 @@ class BaseAgent(ABC):
         """Return local catalog metadata without requesting provider discovery."""
         return list(self.available_models)
 
+    # --- Live catalog ---
+
+    catalog_source: str = "bundled"
+    discovery_error: str = ""
+
+    async def refresh_models(self) -> None:
+        """Replace the bundled catalog with the provider's live list when a key is configured.
+
+        The bundled list stays as an offline fallback and is labelled as such.
+        """
+        from thwip.agents.catalog import fetch_live_models, merge_catalog, supports_live_catalog
+        from thwip.agents.native_common import scrub
+
+        getter = getattr(self, "_get_api_key", None)
+        key = getter() if callable(getter) else None
+        if not key or not supports_live_catalog(self.name):
+            return
+        try:
+            live = await fetch_live_models(self.name, key)
+        except Exception as exc:
+            self.discovery_error = f"Live model list unavailable; showing the bundled fallback. {scrub(str(exc), 160)}"
+            return
+        if live:
+            bundled = type(self).available_models if isinstance(type(self).__dict__.get("available_models"), list) else self.available_models
+            self.available_models = merge_catalog(list(bundled), live)
+            self.catalog_source = "live"
+            self.discovery_error = ""
+
     def has_capability(self, cap: Capability) -> bool:
         """Check if this agent supports a capability."""
         return cap in self.capabilities