coderouter-cli 0.1.0

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.
package/README.md ADDED
@@ -0,0 +1,60 @@
1
+ # coderouter-cli
2
+
3
+ **Route smarter. Ship faster.** CodeRouter is a coding-agent orchestration layer: it classifies each task by its *cognitive shape*, picks the best model from the providers you've enabled, runs edits safely in a git-worktree sandbox, validates the result, and remembers what works on your repo.
4
+
5
+ It works with your existing **Claude Code** / **Codex** CLI, or with any API key — **OpenAI, Anthropic, OpenRouter, DeepSeek, Groq** — via a built-in first-party coding agent.
6
+
7
+ ## Install
8
+
9
+ Requires **Node 24+** (uses Node's built-in SQLite).
10
+
11
+ ```bash
12
+ npm install -g coderouter-cli
13
+ ```
14
+
15
+ Then run:
16
+
17
+ ```bash
18
+ coderouter
19
+ ```
20
+
21
+ That's the whole setup — a prebuilt `ripgrep` is bundled in, and first launch walks you through adding an API key or auto-detects a local Claude Code / Codex CLI.
22
+
23
+ Prefer no global install?
24
+
25
+ ```bash
26
+ npx coderouter-cli
27
+ ```
28
+
29
+ ## Usage
30
+
31
+ ```bash
32
+ coderouter # interactive REPL
33
+ coderouter agent "rename getCwd" # one-shot execution
34
+ coderouter plan "design a cache layer"
35
+ coderouter masterplan "design L1-L5 memory" # research-grade, cited plan
36
+ coderouter debug "tests fail in CI but pass locally"
37
+ coderouter review # review the current diff
38
+ coderouter route "fix typo" # classify + show the chosen route (no run)
39
+ coderouter dashboard # local usage + settings dashboard
40
+ coderouter memory show # inspect persistent memory
41
+ ```
42
+
43
+ In the REPL, type `/` for commands and `@` to reference files. Slash commands include
44
+ `/plan` `/masterplan` `/agent` `/debug` `/review` `/effort low|medium|high|max` `/apply` `/fast` `/clear` `/help` `/exit`.
45
+
46
+ ## How it works
47
+
48
+ - **You stop choosing models.** Each prompt is scored across a cognitive shape (deep reasoning, long context, multi-file taste, adversarial, exploratory) and routed to the right model among the providers you've configured. Set preferred "strong" and "cheap" models in the dashboard if you want to steer it.
49
+ - **Edits are sandboxed.** Every `agent` run executes in a git worktree under `.coderouter/runs/<id>/`, so your repo is never touched mid-run. You get a diff, validators, and a report; changes apply only when you accept them.
50
+ - **It remembers.** Runs, classifications, route stats, and failure patterns persist to `.coderouter/memory.db` so routing improves on your repo over time.
51
+ - **You see why.** Every decision carries a rationale — which classifier fired, which shape axis dominated, which validator failed.
52
+
53
+ ## Links
54
+
55
+ - Repository & full docs: <https://github.com/EfeAcar6431/CodeRouter>
56
+ - Issues: <https://github.com/EfeAcar6431/CodeRouter/issues>
57
+
58
+ ## License
59
+
60
+ MIT