@phi-code-admin/phi-code 0.92.0 → 0.92.1
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/CHANGELOG.md +9 -0
- package/README.md +47 -0
- package/package.json +1 -1
package/CHANGELOG.md
CHANGED
|
@@ -1,5 +1,14 @@
|
|
|
1
1
|
# Changelog
|
|
2
2
|
|
|
3
|
+
## [0.92.1] - 2026-07-12
|
|
4
|
+
|
|
5
|
+
### Docs
|
|
6
|
+
|
|
7
|
+
- README: document the execution-grounded modes (/plan, /debug, /build, /sandbox,
|
|
8
|
+
the sandbox_run oracle, per-phase model routing, the honest n=13 measurement)
|
|
9
|
+
and the one-command Alibaba Cloud Coding Plan setup — three releases of
|
|
10
|
+
features were undocumented.
|
|
11
|
+
|
|
3
12
|
## [0.92.0] - 2026-07-11
|
|
4
13
|
|
|
5
14
|
### Fixed — the two defects the n=13 SWE-bench measurement exposed
|
package/README.md
CHANGED
|
@@ -16,6 +16,7 @@ phi-code runs in four modes: interactive, print or JSON, RPC for process integra
|
|
|
16
16
|
|
|
17
17
|
- [Quick Start](#quick-start)
|
|
18
18
|
- [Providers & Models](#providers--models)
|
|
19
|
+
- [Execution-Grounded Modes (/plan, /debug, /build, /sandbox)](#execution-grounded-modes)
|
|
19
20
|
- [Interactive Mode](#interactive-mode)
|
|
20
21
|
- [Editor](#editor)
|
|
21
22
|
- [Commands](#commands)
|
|
@@ -103,13 +104,59 @@ For each built-in provider, phi maintains a list of tool-capable models, updated
|
|
|
103
104
|
- Xiaomi MiMo Token Plan (China)
|
|
104
105
|
- Xiaomi MiMo Token Plan (Amsterdam)
|
|
105
106
|
- Xiaomi MiMo Token Plan (Singapore)
|
|
107
|
+
- Alibaba Cloud Coding Plan (`qwen3.7-plus`, GLM, Kimi, MiniMax — subscription quota)
|
|
106
108
|
|
|
107
109
|
See [docs/providers.md](docs/providers.md) for detailed setup instructions.
|
|
108
110
|
|
|
111
|
+
**One-command setup for known providers:** `/keys set alibaba-codingplan sk-sp-…` (or `opencode-go <key>`) bootstraps the full provider entry — endpoint, protocol, model list — from just the API key, then validates it with `/keys test <id>`. The live catalog refreshes on session start.
|
|
112
|
+
|
|
109
113
|
**Custom providers & models:** Add providers via `~/.phi/agent/models.json` if they speak a supported API (OpenAI, Anthropic, Google). For custom APIs or OAuth, use extensions. See [docs/models.md](docs/models.md) and [docs/custom-provider.md](docs/custom-provider.md).
|
|
110
114
|
|
|
111
115
|
---
|
|
112
116
|
|
|
117
|
+
## Execution-Grounded Modes
|
|
118
|
+
|
|
119
|
+
Phi's multi-phase modes share one principle, learned the hard way on SWE-bench:
|
|
120
|
+
**execution is the only oracle**. A verdict must come from running real code —
|
|
121
|
+
never from a model reviewing its own output (which we measured approving wrong
|
|
122
|
+
code, even across model families). Full design:
|
|
123
|
+
[docs/design/plan-debug-build.md](docs/design/plan-debug-build.md).
|
|
124
|
+
|
|
125
|
+
| Command | Job | Pipeline |
|
|
126
|
+
|---|---|---|
|
|
127
|
+
| `/plan <spec>` | Plan AND build a project | EXPLORE → PLAN → CODE → TEST → REVIEW |
|
|
128
|
+
| `/debug <failing test \| repro \| description>` | Turn a REAL failure green | REPRODUCE → LOCALIZE → FIX → VERIFY |
|
|
129
|
+
| `/build <spec>` | Build until it runs | EXPLORE → PLAN → CODE → BUILD-VERIFY (run recipe + acceptance + executable red-team) |
|
|
130
|
+
| `/sandbox [status\|prepare\|run <cmd>]` | Inspect the guaranteed environment | — |
|
|
131
|
+
|
|
132
|
+
**The execution sandbox.** Every oracle run goes through the `sandbox_run` tool:
|
|
133
|
+
a per-project Docker container (toolchain auto-detected — node/python/go/rust/ruby
|
|
134
|
+
or your `Dockerfile`; override anything in `.phi/sandbox.json`) that returns the
|
|
135
|
+
REAL exit code. No Docker and nothing to containerize → runs fall back to the
|
|
136
|
+
host; if an environment is demanded but unavailable, phases report `BLOCKED`
|
|
137
|
+
rather than fabricate a pass.
|
|
138
|
+
|
|
139
|
+
**Honesty guarantees** (each one exists because we measured its absence losing):
|
|
140
|
+
- `/debug` refuses to "fix" a failure it cannot reproduce — and a BLOCKED verdict
|
|
141
|
+
gets one second chance on a different model family before the halt.
|
|
142
|
+
- VERIFY requires the reproduction to pass AND the suite not to regress; a
|
|
143
|
+
TIMEOUT counts as a regression (a fix that hangs the suite is worse than none).
|
|
144
|
+
- `/build` reports `SUCCESS` only when a real run meets the acceptance criteria,
|
|
145
|
+
otherwise an honest `PARTIAL` listing what still fails.
|
|
146
|
+
|
|
147
|
+
**Per-phase model routing.** Each phase can run on its own model via
|
|
148
|
+
`~/.phi/agent/routing.json` (e.g. code on `alibaba-codingplan/qwen3.7-plus`,
|
|
149
|
+
verification on a different family so it does not share the coder's blind spot),
|
|
150
|
+
with automatic fallback on transient provider errors.
|
|
151
|
+
|
|
152
|
+
**Measured, honestly:** on a 13-instance SWE-bench-lite slice (official harness),
|
|
153
|
+
a strong single-shot baseline resolved 7/13 vs `/debug` 6/13 at ~2.5× the time —
|
|
154
|
+
with `/debug` producing smaller patches and zero toxic ones. These modes are
|
|
155
|
+
shipped as *tools with honest failure modes*, not as a claimed benchmark win;
|
|
156
|
+
the measurement lives in the repo and is re-runnable.
|
|
157
|
+
|
|
158
|
+
---
|
|
159
|
+
|
|
113
160
|
## Interactive Mode
|
|
114
161
|
|
|
115
162
|
<p align="center"><img src="docs/images/interactive-mode.png" alt="Interactive Mode" width="600"></p>
|