@chanto/cli 0.1.5 → 0.1.7

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/README.md +21 -74
  2. package/package.json +6 -6
package/README.md CHANGED
@@ -6,7 +6,7 @@
6
6
 
7
7
  <p align="center">
8
8
  A fast, lean coding agent for your terminal.
9
- Any model, local or hosted, including your cloud's:
9
+ Local and cloud models, including:
10
10
  Amazon Bedrock, Google Vertex AI and Azure.
11
11
  </p>
12
12
 
@@ -14,17 +14,16 @@
14
14
  <img alt="Status: alpha" src="https://img.shields.io/badge/status-alpha-f59e0b">
15
15
  <a href="https://scorecard.dev/viewer/?uri=github.com/chanto-sh/chanto"><img alt="OpenSSF Scorecard" src="https://api.scorecard.dev/projects/github.com/chanto-sh/chanto/badge"></a>
16
16
  <a href="https://www.bestpractices.dev/projects/15833"><img alt="OpenSSF Best Practices" src="https://www.bestpractices.dev/projects/15833/badge"></a>
17
+ <a href="https://codecov.io/gh/chanto-sh/chanto"><img alt="Coverage" src="https://codecov.io/gh/chanto-sh/chanto/graph/badge.svg"></a>
17
18
  <img alt="License" src="https://img.shields.io/badge/license-Apache--2.0-blue">
18
19
  <img alt="Platforms" src="https://img.shields.io/badge/platforms-macOS%20·%20Linux%20·%20Windows-0078D6">
19
20
  </p>
20
21
 
21
22
  > [!NOTE]
22
23
  > **Chanto is in alpha:** expect rough edges and breaking changes before 1.0.
23
- > It's built by one developer with a lot of help from AI coding agents; CI runs
24
- > the tests on Linux, macOS and Windows, and the [security model](docs/security-model.md)
25
- > explains how it protects you. [Issues](https://github.com/chanto-sh/chanto/issues)
26
- > and feedback are welcome. Switching from another agent? Chanto reads the config
27
- > you already have.
24
+ > CI tests Linux, macOS and Windows. Read the [security model](docs/security-model.md)
25
+ > for the protections and their limits. [Issues](https://github.com/chanto-sh/chanto/issues)
26
+ > and feedback are welcome.
28
27
 
29
28
  <p align="center">
30
29
  <a href="docs/reference.md"><b>Docs</b></a> ·
@@ -53,7 +52,9 @@ cd your-project
53
52
  chanto
54
53
  ```
55
54
 
56
- Chanto looks for what you already have: API keys in your environment, your Codex config, a running Ollama. Then it asks you to pick a model. No keys? Pick one of the free options: OpenRouter's free models, Gemini's free tier, or a model on your own machine. [Hosted models](docs/reference.md#hosted-models) covers every provider, including your ChatGPT plan, Amazon Bedrock, Google Vertex AI and Azure.
55
+ Chanto detects API keys, compatible configuration and local model servers, then asks you to pick a model. Use an API provider, your ChatGPT plan, or a local model through Ollama or an OpenAI-compatible server.
56
+
57
+ Run `chanto login chatgpt` for your ChatGPT plan. Subscription coding plans such as Kimi Code (`chanto login kimi-coding`) and Z.ai (`chanto login zai`) use their own plan keys. Login saves credentials; `/model` or `/models` selects the model. See [providers and login](docs/reference.md#hosted-models).
57
58
 
58
59
  Then just say what you want, like "fix the failing test in auth". Chanto asks before editing files or running risky commands, unless you allow them.
59
60
 
@@ -91,41 +92,17 @@ In a session:
91
92
 
92
93
  | | |
93
94
  |---|---|
94
- | **Fast and lean** | One native binary that starts instantly and stays light on CPU and memory while it works ([benchmarks](#benchmarks)) |
95
- | **Any model** | Free to start (OpenRouter's free models, Gemini's free tier, or a local model); Ollama, llama.cpp, vLLM and other OpenAI-compatible servers; OpenAI, Anthropic, OpenRouter, DeepSeek and more; your ChatGPT login; Amazon Bedrock, Google Vertex AI and Azure with your cloud credentials |
96
- | **Safe by default** | Shell commands run in a sandbox on macOS and Linux (on Windows it confines writes only: reads and network aren't restricted), with optional network domain allowlists on macOS and Linux; risky actions need your OK, and auto mode screens commands for you |
97
- | **Undo** | `/rewind` restores files (including what shell commands changed) and the conversation to any earlier prompt |
98
- | **Outcome mode** | Agree on what "done" means first, then review results instead of code |
99
- | **Background runs** | `chanto run --bg "<task>"` (or `&task` in the TUI) works in its own git worktree and branch while you keep going; `chanto runs`, `attach` to answer prompts, `apply` to bring the result over; optional per-run Docker/Podman container for its commands (`[background] isolate = "docker"`, no network or an allowlist, CPU/memory limits) |
100
- | **Checks its work** | Runs the tests related to its change before saying it's done (the full suite when shared setup changed, or with `--verify-full`), tells failures it caused from ones that were already there, and won't weaken tests to pass |
101
- | **Finds its way in big repos** | `RepoMap` gives a ranked, token-budgeted map of the project's symbols; `SemanticSearch` finds code by meaning through `[embeddings]` (a local Ollama model by default). Both are read-only, so they run without a prompt |
102
- | **Plays well with others** | MCP, hooks, and Claude Code plugin marketplaces (`chanto plugin`) |
103
-
104
- <details>
105
- <summary><b>Everything else</b>: goals, teams, scheduled jobs, memory, voice and more</summary>
106
-
107
- | | |
108
- |---|---|
109
- | **Goals** | `/goal <objective>` keeps working across turns and context windows until the agreed, locked criteria pass; stops on limits or no progress, and asks when blocked |
110
- | **Teams** | `/team <task>` splits work into parallel sub-tasks in their own worktrees, reviews each one, merges them onto a team branch and runs your checks |
111
- | **Costs and diagnostics** | USD totals, session/daily budgets, a spend guard that estimates fan-out before it starts and explains costly loops, redacted daily logs, opt-in OTLP telemetry, `/feedback`, and opt-in error reports |
112
- | **Keeps secrets out** | Hides API keys and `.env` values from the model and logs |
113
- | **Durable memory** | Remembers user preferences and project facts across sessions, with visible saves and undo |
114
- | **Local models** | Tuned for small local models, and won't run your machine out of memory |
115
- | **Easy to install** | One-line installers that verify checksums, optional daily update notices, and `chanto upgrade` |
116
- | **Chanto cloud** | `chanto login cloud`, then models as `chanto/<model>` and agents that run on a schedule, on GitHub events or from CI: `chanto agents create`, `chanto agents run --wait`, `chanto cloud run --agent <name>` with `CHANTO_CLOUD_TOKEN` ([docs](docs/reference.md#chanto-cloud)) |
117
- | **Scheduled jobs** | `chanto runner`: cron, interval, webhook and CI triggers, each an isolated background run with budgets, verification and notifications |
118
- | **Drive another agent** | `chanto acp-client <command>` runs any ACP agent (Claude Code via `claude-agent-acp`, OpenCode, …) with Chanto's UI, permissions and your MCP servers; the agent can run commands through Chanto's terminals, its modes, settings and sign-in work through `/agent-mode`, `/agent-config` and `/agent-login`, and its own slash commands are passed through ([docs](docs/reference.md#driving-another-agent-acp-client)) |
119
- | **Learns from you** | Suggests lessons from your corrections, only applied when you approve |
120
- | **Customizable** | Ask `/customize` or `chanto customize` to add commands, skills, hooks and more; `/reload` applies extensions |
121
- | **Computer use** | Opt-in (`--computer`) desktop control beyond the browser: screenshots, clicks, typing and apps on macOS (built-in tools only) and Linux (X11, or Wayland with grim/GNOME/KDE screenshots and ydotool); every action needs your approval, and a status-line indicator shows it's on |
122
- | **Remote runs** | `chanto serve --runs` on your own server, then `chanto run --bg --on <server>`: the code goes as a git bundle or through your git remote, and the results come back the same way |
123
- | **Sessions** | Resume from any directory, fork current or past conversations, browse session trees, share redacted standalone HTML, and import sessions others share with you (`chanto import-session`) |
124
- | **Voice input** | ctrl+space or `/voice` in the TUI: dictate a prompt, transcribed locally with whisper.cpp (or a hosted endpoint you configure) and placed in the prompt for review ([docs](docs/reference.md#voice-input)) |
125
- | **Your terminal, your way** | A full-screen view, or `--inline` to print into your terminal's own scrollback (native scroll, search and copy); Markdown renders as it streams; `/effort` sets how hard the model thinks |
126
-
127
- Full list in the [docs](docs/reference.md).
128
- </details>
95
+ | **Local and cloud models** | OpenAI, Anthropic, Gemini, DeepSeek, OpenRouter and more; ChatGPT and coding plans; Bedrock, Vertex AI and Azure; local OpenAI-compatible servers |
96
+ | **Permission controls** | Ask before risky actions, or use plan, accept-edits and auto modes. Shell sandboxing covers macOS and Linux; Windows restricts writes, with read/network limits described in the [security model](docs/security-model.md) |
97
+ | **Verification** | `--verify` adds a project-check gate before completion; `--verify-full` runs the full suite. The test-weakening guard also runs on ordinary turns ([reference](docs/reference.md#verified-mode-guardrails)) |
98
+ | **Sessions and undo** | Resume, fork and rewind conversations with file changes; share redacted transcripts |
99
+ | **Background work** | Run tasks in separate worktrees, monitor jobs, and apply their results while continuing your session |
100
+ | **Goals and teams** | Keep working toward agreed criteria or split a task across agents with checks and limits |
101
+ | **Extensions** | MCP tools, hooks, skills, custom commands and compatible plugin marketplaces |
102
+ | **Usage and costs** | Session token/cost totals, budgets, and provider-reported limits with low-quota warnings |
103
+
104
+ The [reference guide](docs/reference.md) also covers memory, scheduled jobs,
105
+ voice, remote runs, optional computer control and using another agent through ACP.
129
106
 
130
107
  <details>
131
108
  <summary><b>Compatibility: the existing Claude Code and Codex setup it reads</b></summary>
@@ -142,36 +119,6 @@ Project settings that could run code are ignored until you trust the folder.
142
119
 
143
120
  ---
144
121
 
145
- ## What's different
146
-
147
- Chanto is built around one rule: finishing means evidence, not a confident answer.
148
-
149
- - **A finish it has to earn.** Before accepting "done", Chanto runs the project's tests and sends failures back to the model. The final answer says what actually passed and how much of the suite that was.
150
- - **Tests are protected.** Existing tests are snapshotted each turn. Deleting them, weakening assertions or adding skips is caught and rejected: fix the code, not the test.
151
- - **Stopping is a feature.** Goals and teams keep working until agreed criteria pass, stop on limits or no progress, and ask when blocked instead of spinning.
152
- - **Local-first.** One Rust binary that talks to local or hosted models, reads existing agent config one way, and needs no account to start.
153
-
154
- ---
155
-
156
- ## How it works
157
-
158
- ```mermaid
159
- flowchart LR
160
- you(["You"]) --> agent["Agent"]
161
- agent <--> model["Model<br/><i>local or hosted</i>"]
162
- agent --> tools["Tools<br/><i>files · shell · web · MCP</i>"]
163
- agent --> check{{"Checks<br/><i>tests pass?</i>"}}
164
- check -->|no| agent
165
- check -->|yes| done(["Done"])
166
-
167
- classDef ok fill:#10b98118,stroke:#10b981;
168
- class check,done ok
169
- ```
170
-
171
- Every action goes through a permission check first, and secrets are masked before anything is sent to the model.
172
-
173
- ---
174
-
175
122
  ## Benchmarks
176
123
 
177
124
  Two things are measured, each with the same rules for every agent:
@@ -183,14 +130,14 @@ Results, versions and caveats are in [RESULTS.md](https://github.com/chanto-sh/b
183
130
 
184
131
  ```sh
185
132
  git clone https://github.com/chanto-sh/benchmarks.git && cd benchmarks
186
- python3 reproduce.py --quick # 6 native agents in a few minutes; drop --quick for all 18
133
+ python3 reproduce.py --quick # smaller selection; omit --quick for the full comparison
187
134
  ```
188
135
 
189
136
  ---
190
137
 
191
138
  ## Development
192
139
 
193
- `src/` is the agent, `proto/` the wire types shared with the hosted service, `classifier/` auto mode's command classifier, and `docs/` the reference. It needs Rust 1.95 or newer:
140
+ `src/` contains the agent, `proto/` shared wire types, and `docs/` the reference. Building requires Rust 1.95 or newer:
194
141
 
195
142
  ```sh
196
143
  cargo build # add --release for target/release/chanto
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@chanto/cli",
3
- "version": "0.1.5",
3
+ "version": "0.1.7",
4
4
  "description": "A fast, lean coding agent for your terminal",
5
5
  "license": "Apache-2.0",
6
6
  "homepage": "https://chanto.sh",
@@ -19,10 +19,10 @@
19
19
  "node": ">=18"
20
20
  },
21
21
  "optionalDependencies": {
22
- "@chanto/cli-darwin-arm64": "0.1.5",
23
- "@chanto/cli-darwin-x64": "0.1.5",
24
- "@chanto/cli-linux-arm64": "0.1.5",
25
- "@chanto/cli-linux-x64": "0.1.5",
26
- "@chanto/cli-win32-x64": "0.1.5"
22
+ "@chanto/cli-darwin-arm64": "0.1.7",
23
+ "@chanto/cli-darwin-x64": "0.1.7",
24
+ "@chanto/cli-linux-arm64": "0.1.7",
25
+ "@chanto/cli-linux-x64": "0.1.7",
26
+ "@chanto/cli-win32-x64": "0.1.7"
27
27
  }
28
28
  }