tau-coding-agent 0.6.0__tar.gz → 0.6.2__tar.gz

tau_coding_agent-0.6.2/PKG-INFO

@@ -0,0 +1,213 @@
+Metadata-Version: 2.4
+Name: tau-coding-agent
+Version: 0.6.2
+Summary: A self-extensible agent CLI with terminal UI, multi-provider LLM support, session management with branching, and a plugin system for tools and commands
+Requires-Python: >=3.12
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: anthropic==0.109.1
+Requires-Dist: certifi==2026.5.20
+Requires-Dist: click==8.4.1
+Requires-Dist: filelock==3.29.3
+Requires-Dist: google-genai==2.8.0
+Requires-Dist: httpx==0.28.1
+Requires-Dist: mistralai==2.4.9
+Requires-Dist: ollama==0.6.2
+Requires-Dist: openai==2.41.1
+Requires-Dist: packaging==26.2
+Requires-Dist: pillow==12.2.0
+Requires-Dist: pydantic==2.13.4
+Requires-Dist: python-dotenv==1.2.2
+Requires-Dist: uuid7==0.1.0
+Requires-Dist: websockets==16.0
+Requires-Dist: mistletoe==1.5.1
+Requires-Dist: pyyaml==6.0.2
+Requires-Dist: pygments==2.20.0
+Requires-Dist: pathspec==1.1.1
+Requires-Dist: posthog==7.21.3
+Requires-Dist: pylatexenc==2.10
+Requires-Dist: gitpython==3.1.50
+Requires-Dist: grapheme==0.6.0
+Requires-Dist: arrow==1.4.0
+Requires-Dist: kiwisolver==1.5.0
+Requires-Dist: tiktoken==0.13.0
+Provides-Extra: tools
+Requires-Dist: ripgrep>=14.0.0; extra == "tools"
+Requires-Dist: ast-grep-cli>=0.44.0; extra == "tools"
+Dynamic: license-file
+
+# Tau (τ)
+
+Tau is a Python agent framework and terminal coding assistant. It combines an
+interactive terminal UI, multiple model providers, persistent sessions, tool
+execution, and an extension system in one package.
+
+<p align="center">
+  <img src="assets/tui.jpeg" alt="Tau interactive terminal interface" width="700">
+</p>
+
+## Quick start
+
+Tau requires Python 3.12 or later.
+
+```bash
+pip install tau-coding-agent
+```
+
+Set a provider API key and start Tau:
+
+```bash
+export ANTHROPIC_API_KEY=sk-ant-...
+tau
+```
+
+Google AI Studio works through the `google` provider:
+
+```bash
+export GOOGLE_API_KEY=...
+tau --model google/gemini-2.5-flash
+```
+
+You can also run `/login` inside Tau to save provider credentials.
+
+Then ask Tau to work in the current directory:
+
+```text
+Explain this repository, run its tests, and fix any failures.
+```
+
+## Common workflows
+
+```bash
+tau                                      # Start an interactive session
+tau --resume                             # Resume the latest session
+tau --model claude-sonnet-4-6            # Start with a specific model
+tau --print "Summarize this repository"  # Run once and print the result
+tau --mode json "Summarize this repo"    # Emit structured JSON events
+tau --mode rpc                           # Start JSON-RPC mode for IDE clients
+```
+
+Inside an interactive session:
+
+```text
+/model       Choose a model
+/resume      Resume another session
+/tree        Navigate session branches
+/compact     Compact a long conversation
+/theme       Change the terminal theme
+/login       Save provider credentials
+/help        Show commands and shortcuts
+```
+
+See the [CLI reference](docs/cli-reference.md) for every option and command.
+
+## What Tau provides
+
+- **Interactive terminal UI** with multiline editing, searchable pickers,
+  syntax highlighting, Markdown, and terminal-readable LaTeX math.
+- **Multiple model providers**, including Anthropic, OpenAI, Google Gemini,
+  Mistral, Ollama, Groq, xAI, Bedrock, OpenRouter, and others.
+- **Persistent session trees** with resume, fork, clone, branch navigation,
+  summarization, and automatic context compaction.
+- **Built-in tools** for terminal commands, file operations, globbing, and
+  search. Long-running terminal commands stream into one persistent output
+  block.
+- **Media support** for images, audio, video, and text files through file
+  references, clipboard input, and the Python API.
+- **Speech APIs** for text-to-speech and speech-to-text, including word or
+  segment timestamps when supported by the selected provider.
+- **Extensibility** through custom tools, slash commands, hooks, themes,
+  skills, prompts, and in-memory Python extensions.
+- **Embedding and integration** through the Python API, JSON event mode, and
+  bidirectional JSON-RPC.
+
+## Referencing files
+
+Type `@` in the interactive editor to search for a project file:
+
+```text
+Review @src/service.py and add tests for its error handling.
+```
+
+For one-shot execution, attach a file explicitly:
+
+```bash
+tau --print --prompt "Explain this file" --file src/service.py
+```
+
+Tau also discovers project instructions from `AGENTS.md` and `CLAUDE.md`.
+See [Project Context Files](docs/project-context.md) for trust and discovery
+behavior.
+
+## Authentication and configuration
+
+Tau resolves provider credentials in this order:
+
+1. A programmatic runtime override
+2. A credential saved in `~/.tau/auth.json` (including keys saved by `/login`)
+3. A provider environment variable such as `ANTHROPIC_API_KEY`,
+   `OPENAI_API_KEY`, and `GOOGLE_API_KEY`
+
+Settings are merged in this order:
+
+1. Built-in defaults
+2. `~/.tau/settings.json`
+3. `.tau/settings.json`
+4. Environment variables
+5. Command-line options
+
+See [Authentication](docs/auth.md), [Installation](docs/installation.md), and
+[Inference Providers](docs/inference-providers.md) for provider-specific
+setup.
+
+## Documentation
+
+- [Quickstart](docs/quickstart.md) — First session in five minutes
+- [Usage](docs/usage.md) — Interactive workflows and commands
+- [CLI Reference](docs/cli-reference.md) — Command-line options and modes
+- [Inference Providers](docs/inference-providers.md) — Providers and speech timestamps
+- [Sessions](docs/sessions.md) — Persistence, branching, and compaction
+- [Tools](docs/tools.md) — Built-in and custom tools
+- [Extensions](docs/extensions.md) — Tools, commands, hooks, and plugins
+- [Terminal UI](docs/tui.md) — Rendering, Markdown, math, and components
+- [Python API](docs/python-api.md) — Embed Tau in another application
+- [Architecture](docs/architecture.md) — Internal design and data flow
+
+The complete documentation index is available at [docs/index.md](docs/index.md).
+
+## Install from source
+
+```bash
+git clone https://github.com/Jeomon/Tau.git
+cd Tau
+pip install -e .
+tau
+```
+
+## Security
+
+Tau executes enabled tools with the operating-system permissions of the process
+that launched it. Review project instructions and commands before approving
+work in untrusted repositories. Use a container or external sandbox when
+stronger isolation is required.
+
+Dependency versions are pinned and recorded in `uv.lock`. See
+[SECURITY.md](SECURITY.md) for vulnerability reporting and supply-chain
+practices.
+
+## Development
+
+```bash
+mypy tau/
+pyright tau/
+ruff check tau/
+ruff format tau/
+python -m pytest
+```
+
+See [Development Setup](docs/development.md) and
+[Contributing](CONTRIBUTING.md).
+
+## License
+
+Tau is licensed under the [MIT License](LICENSE).

tau_coding_agent-0.6.2/README.md

@@ -0,0 +1,175 @@
+# Tau (τ)
+
+Tau is a Python agent framework and terminal coding assistant. It combines an
+interactive terminal UI, multiple model providers, persistent sessions, tool
+execution, and an extension system in one package.
+
+<p align="center">
+  <img src="assets/tui.jpeg" alt="Tau interactive terminal interface" width="700">
+</p>
+
+## Quick start
+
+Tau requires Python 3.12 or later.
+
+```bash
+pip install tau-coding-agent
+```
+
+Set a provider API key and start Tau:
+
+```bash
+export ANTHROPIC_API_KEY=sk-ant-...
+tau
+```
+
+Google AI Studio works through the `google` provider:
+
+```bash
+export GOOGLE_API_KEY=...
+tau --model google/gemini-2.5-flash
+```
+
+You can also run `/login` inside Tau to save provider credentials.
+
+Then ask Tau to work in the current directory:
+
+```text
+Explain this repository, run its tests, and fix any failures.
+```
+
+## Common workflows
+
+```bash
+tau                                      # Start an interactive session
+tau --resume                             # Resume the latest session
+tau --model claude-sonnet-4-6            # Start with a specific model
+tau --print "Summarize this repository"  # Run once and print the result
+tau --mode json "Summarize this repo"    # Emit structured JSON events
+tau --mode rpc                           # Start JSON-RPC mode for IDE clients
+```
+
+Inside an interactive session:
+
+```text
+/model       Choose a model
+/resume      Resume another session
+/tree        Navigate session branches
+/compact     Compact a long conversation
+/theme       Change the terminal theme
+/login       Save provider credentials
+/help        Show commands and shortcuts
+```
+
+See the [CLI reference](docs/cli-reference.md) for every option and command.
+
+## What Tau provides
+
+- **Interactive terminal UI** with multiline editing, searchable pickers,
+  syntax highlighting, Markdown, and terminal-readable LaTeX math.
+- **Multiple model providers**, including Anthropic, OpenAI, Google Gemini,
+  Mistral, Ollama, Groq, xAI, Bedrock, OpenRouter, and others.
+- **Persistent session trees** with resume, fork, clone, branch navigation,
+  summarization, and automatic context compaction.
+- **Built-in tools** for terminal commands, file operations, globbing, and
+  search. Long-running terminal commands stream into one persistent output
+  block.
+- **Media support** for images, audio, video, and text files through file
+  references, clipboard input, and the Python API.
+- **Speech APIs** for text-to-speech and speech-to-text, including word or
+  segment timestamps when supported by the selected provider.
+- **Extensibility** through custom tools, slash commands, hooks, themes,
+  skills, prompts, and in-memory Python extensions.
+- **Embedding and integration** through the Python API, JSON event mode, and
+  bidirectional JSON-RPC.
+
+## Referencing files
+
+Type `@` in the interactive editor to search for a project file:
+
+```text
+Review @src/service.py and add tests for its error handling.
+```
+
+For one-shot execution, attach a file explicitly:
+
+```bash
+tau --print --prompt "Explain this file" --file src/service.py
+```
+
+Tau also discovers project instructions from `AGENTS.md` and `CLAUDE.md`.
+See [Project Context Files](docs/project-context.md) for trust and discovery
+behavior.
+
+## Authentication and configuration
+
+Tau resolves provider credentials in this order:
+
+1. A programmatic runtime override
+2. A credential saved in `~/.tau/auth.json` (including keys saved by `/login`)
+3. A provider environment variable such as `ANTHROPIC_API_KEY`,
+   `OPENAI_API_KEY`, and `GOOGLE_API_KEY`
+
+Settings are merged in this order:
+
+1. Built-in defaults
+2. `~/.tau/settings.json`
+3. `.tau/settings.json`
+4. Environment variables
+5. Command-line options
+
+See [Authentication](docs/auth.md), [Installation](docs/installation.md), and
+[Inference Providers](docs/inference-providers.md) for provider-specific
+setup.
+
+## Documentation
+
+- [Quickstart](docs/quickstart.md) — First session in five minutes
+- [Usage](docs/usage.md) — Interactive workflows and commands
+- [CLI Reference](docs/cli-reference.md) — Command-line options and modes
+- [Inference Providers](docs/inference-providers.md) — Providers and speech timestamps
+- [Sessions](docs/sessions.md) — Persistence, branching, and compaction
+- [Tools](docs/tools.md) — Built-in and custom tools
+- [Extensions](docs/extensions.md) — Tools, commands, hooks, and plugins
+- [Terminal UI](docs/tui.md) — Rendering, Markdown, math, and components
+- [Python API](docs/python-api.md) — Embed Tau in another application
+- [Architecture](docs/architecture.md) — Internal design and data flow
+
+The complete documentation index is available at [docs/index.md](docs/index.md).
+
+## Install from source
+
+```bash
+git clone https://github.com/Jeomon/Tau.git
+cd Tau
+pip install -e .
+tau
+```
+
+## Security
+
+Tau executes enabled tools with the operating-system permissions of the process
+that launched it. Review project instructions and commands before approving
+work in untrusted repositories. Use a container or external sandbox when
+stronger isolation is required.
+
+Dependency versions are pinned and recorded in `uv.lock`. See
+[SECURITY.md](SECURITY.md) for vulnerability reporting and supply-chain
+practices.
+
+## Development
+
+```bash
+mypy tau/
+pyright tau/
+ruff check tau/
+ruff format tau/
+python -m pytest
+```
+
+See [Development Setup](docs/development.md) and
+[Contributing](CONTRIBUTING.md).
+
+## License
+
+Tau is licensed under the [MIT License](LICENSE).

{tau_coding_agent-0.6.0 → tau_coding_agent-0.6.2}/docs/installation.md

@@ -61,9 +61,12 @@ Get your API key from [OpenAI's platform](https://platform.openai.com/account/ap
 
 ```bash
 export GOOGLE_API_KEY=...
+tau --model google/gemini-2.5-flash
 ```
 
-Create a key at [Google AI Studio](https://aistudio.google.com).
+Create a key at [Google AI Studio](https://aistudio.google.com). The `google`
+provider uses the Gemini Developer API and is separate from the Google Vertex
+AI provider.
 
 ### Mistral AI
 
@@ -144,7 +147,7 @@ When you start Tau, it will load your models. Press `/model` to see all availabl
 To remove Tau:
 
 ```bash
-pip uninstall tau
+pip uninstall tau-coding-agent
 ```
 
 This removes the tau command but leaves configuration and session data in `~/.tau/`.

{tau_coding_agent-0.6.0 → tau_coding_agent-0.6.2}/docs/quickstart.md

@@ -33,6 +33,7 @@ export OPENAI_API_KEY=sk-...
 
 ```bash
 export GOOGLE_API_KEY=...
+tau --model google/gemini-2.5-flash
 ```
 
 ### Other Providers
@@ -66,6 +67,10 @@ Summarize this repository and tell me how to run its checks.
 
 The agent will read files, run commands, and respond with a summary.
 
+Markdown responses support terminal-readable inline and display LaTeX math.
+Long-running terminal tools update their existing output block while streaming
+instead of printing a new block for every chunk.
+
 ### Media Support
 
 Tau can process multiple file types. Reference them with `@` during input:

{tau_coding_agent-0.6.0 → tau_coding_agent-0.6.2}/docs/tools.md

@@ -10,8 +10,8 @@ Tau ships with seven built-in tools covering file I/O, search, and shell executi
 
 Read a UTF-8 text file. Invalid byte sequences are replaced during decoding.
 Every returned line is prefixed with a content-based hashline anchor in the form
-`<line>:<hash>|<content>`. Duplicate content, including blank lines, can share a
-hash; the line number acts as a proximity hint when `edit` resolves an anchor.
+`<line>:<hash>|<content>`. Duplicate content, including blank lines, receives
+distinct anchors so `edit` can target the exact displayed line.
 
 | Parameter | Type | Required | Default | Description |
 |-----------|------|----------|---------|-------------|
@@ -37,6 +37,8 @@ used as a proximity hint and the closest occurrence is selected. An empty
 `new_content` deletes the selected range. Because the complete file is rewritten,
 an edit may normalize line endings. Edit-result diffs also display hashline anchors:
 removed lines use their old hashes, while added and context lines use current hashes.
+If an anchor hash is stale or invalid, the model-visible error includes current
+nearby hashline-anchored file content and asks the agent to re-read before retrying.
 Edit diffs always show every changed line. By default they include three unchanged
 lines around each change and collapse larger unchanged gaps into `… (+N lines)`;
 Ctrl+O expands those gaps to show the complete file context.

{tau_coding_agent-0.6.0 → tau_coding_agent-0.6.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "tau-coding-agent"
-version = "0.6.0"
+version = "0.6.2"
 description = "A self-extensible agent CLI with terminal UI, multi-provider LLM support, session management with branching, and a plugin system for tools and commands"
 readme = "README.md"
 requires-python = ">=3.12"
@@ -34,6 +34,7 @@ dependencies = [
     "grapheme==0.6.0",
     "arrow==1.4.0",
     "kiwisolver==1.5.0",
+    "tiktoken==0.13.0",
 ]
 
 [project.scripts]

{tau_coding_agent-0.6.0 → tau_coding_agent-0.6.2}/tau/agent/prompt/builder.py

@@ -146,6 +146,18 @@ _GENERAL_GUIDELINES = [
         "Prioritize accuracy over agreement — investigate before confirming, and "
         "disagree when the evidence calls for it."
     ),
+    (
+        "Verify before reporting a task done: run the tests, build, or command that "
+        "exercises the change. Don't present untested, partial, or stubbed-out work as complete."
+    ),
+    (
+        "When fixing a bug, reproduce it before changing code. If the same fix attempt "
+        "fails twice, stop and investigate the root cause instead of retrying it."
+    ),
+    (
+        "Treat git history as shared: never force-push, amend a commit that isn't yours, "
+        "or skip hooks (--no-verify) without the user's explicit go-ahead."
+    ),
 ]
 
 _PRECEDENCE_GUIDELINES = """\

{tau_coding_agent-0.6.0 → tau_coding_agent-0.6.2}/tau/agent/service.py

@@ -20,6 +20,7 @@ from tau.message.utils import strip_unusable_trailing_assistant
 from tau.session.compaction import CompactionSettings
 from tau.session.utils import to_llm_messages as _to_llm_messages
 from tau.tool.types import ToolInvocation, ToolResult
+from tau.utils.format import human_size as _fmt_size
 
 _log = logging.getLogger(__name__)
 
@@ -33,14 +34,6 @@ _TOOL_CAP_LINES = 2000  # DEFAULT_MAX_LINES
 _TOOL_LINE_CAP_BYTES = 2 * 1024  # 2 KB — max bytes for a single line
 
 
-def _fmt_size(n: int) -> str:
-    if n < 1024:
-        return f"{n}B"
-    if n < 1024 * 1024:
-        return f"{n / 1024:.1f}KB"
-    return f"{n / (1024 * 1024):.1f}MB"
-
-
 if TYPE_CHECKING:
     from tau.engine.service import Engine
     from tau.runtime.service import Runtime
@@ -509,6 +502,31 @@ class Agent:
             self._record_compaction_failure("Auto-compaction failed")
             return False
 
+    def _estimate_indicates_overflow(self) -> bool:
+        """Numeric fallback for overflow detection, independent of error text.
+
+        A failed request never gets a provider-reported usage back, so there's
+        no "provider said input tokens exceeded the window" signal to check —
+        only Tau's own pre-send estimate for what was just sent. Some providers
+        reject an over-window request with phrasing that matches none of
+        _CONTEXT_OVERFLOW_PATTERNS (tau/inference/utils.py) — NVIDIA's
+        "max_tokens must be at least 1, got -128" was one instance, where the
+        gateway computed context_window - prompt_tokens server-side and
+        rejected the negative result. If Tau's own estimate for the request
+        that just failed already reached the model's context window, treat it
+        as overflow regardless of how the provider phrased the rejection.
+        """
+        from tau.session.compaction import estimate_context_tokens
+
+        if self._context_window <= 0:
+            return False
+        session_ctx = self._session_manager.build_session_context()
+        llm_messages = _to_llm_messages(session_ctx.messages)
+        usage = estimate_context_tokens(
+            llm_messages, system_prompt=self._system_prompt, tools=self._engine.tools
+        )
+        return usage.tokens >= self._context_window
+
     async def _try_overflow_recovery(self) -> bool:
         """If the last turn died with a context-overflow error, compact once and signal a retry.
 
@@ -524,7 +542,9 @@ class Agent:
             return False
 
         last = self._session_manager.find_last_assistant_message()
-        if last is None or last.error_kind != ErrorKind.CONTEXT_OVERFLOW:
+        if last is None:
+            return False
+        if last.error_kind != ErrorKind.CONTEXT_OVERFLOW and not self._estimate_indicates_overflow():
             return False
 
         # Model-switch guard: the overflow error is from a different model if it predates

{tau_coding_agent-0.6.0 → tau_coding_agent-0.6.2}/tau/auth/storage.py

@@ -2,8 +2,6 @@ from abc import ABC, abstractmethod
 from collections.abc import Awaitable, Callable
 from pathlib import Path
 
-from filelock import FileLock
-
 from tau.auth.types import LockResult
 
 
@@ -45,6 +43,8 @@ class FileAuthStorage(AuthStorage):
 
     def with_lock(self, fn: Callable[[str | None], LockResult]) -> LockResult:
         """Execute fn with exclusive access to storage."""
+        from filelock import FileLock
+
         with FileLock(self.lock_path):
             current = (
                 self.store_path.read_text(encoding="utf-8") if self.store_path.exists() else None
@@ -59,6 +59,8 @@ class FileAuthStorage(AuthStorage):
         self, fn: Callable[[str | None], Awaitable[LockResult]]
     ) -> LockResult:
         """Execute async fn with exclusive access to storage."""
+        from filelock import FileLock
+
         with FileLock(self.lock_path):
             current = (
                 self.store_path.read_text(encoding="utf-8") if self.store_path.exists() else None