sugarai 3.1.1.dev4__tar.gz → 3.3.1__tar.gz
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.
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/PKG-INFO +120 -4
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/README.md +114 -2
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/pyproject.toml +7 -2
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/agent/base.py +72 -56
- sugarai-3.3.1/sugar/config/__init__.py +6 -0
- sugarai-3.3.1/sugar/config/issue_responder_config.py +143 -0
- sugarai-3.3.1/sugar/config/prompt_config.py +135 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/core/loop.py +20 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/discovery/github_watcher.py +154 -2
- sugarai-3.3.1/sugar/executor/agent_sdk_executor.py +441 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/main.py +174 -1
- sugarai-3.3.1/sugar/mcp/__init__.py +34 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/profiles/issue_responder.py +16 -1
- sugarai-3.3.1/sugar/ralph/__init__.py +19 -0
- sugarai-3.3.1/sugar/ralph/config.py +75 -0
- sugarai-3.3.1/sugar/ralph/profile.py +365 -0
- sugarai-3.3.1/sugar/ralph/validator.py +293 -0
- sugarai-3.3.1/sugar/storage/__init__.py +5 -0
- sugarai-3.3.1/sugar/storage/issue_response_manager.py +258 -0
- sugarai-3.3.1/sugar/triage/__init__.py +43 -0
- sugarai-3.3.1/sugar/triage/capability_scanner.py +752 -0
- sugarai-3.3.1/sugar/triage/complexity_analyzer.py +739 -0
- sugarai-3.3.1/sugar/triage/triage_analyzer.py +440 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugarai.egg-info/PKG-INFO +120 -4
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugarai.egg-info/SOURCES.txt +19 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugarai.egg-info/requires.txt +5 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_agent.py +24 -1
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_cli.py +105 -0
- sugarai-3.3.1/tests/test_complexity_analyzer.py +634 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_core_loop.py +48 -0
- sugarai-3.3.1/tests/test_github_client.py +561 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_goose_integration.py +3 -2
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_issue_responder_cli.py +36 -7
- sugarai-3.3.1/tests/test_issue_response_integration.py +1042 -0
- sugarai-3.3.1/tests/test_mcp_integration.py +498 -0
- sugarai-3.3.1/tests/test_mcp_server.py +253 -0
- sugarai-3.3.1/tests/test_prompt_config.py +359 -0
- sugarai-3.3.1/tests/test_ralph.py +410 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_subagent_manager.py +13 -9
- sugarai-3.1.1.dev4/sugar/executor/agent_sdk_executor.py +0 -223
- sugarai-3.1.1.dev4/sugar/mcp/__init__.py +0 -11
- sugarai-3.1.1.dev4/sugar/utils/__init__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/LICENSE +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/setup.cfg +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/__init__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/__version__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/agent/__init__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/agent/hooks.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/agent/subagent_manager.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/agent/tools.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/billing/__init__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/billing/api_keys.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/billing/tiers.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/billing/usage.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/core/__init__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/discovery/__init__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/discovery/code_quality.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/discovery/error_monitor.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/discovery/test_coverage.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/executor/__init__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/executor/base.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/executor/claude_wrapper.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/executor/structured_request.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/integrations/__init__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/integrations/github.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/learning/__init__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/learning/adaptive_scheduler.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/learning/feedback_processor.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/mcp/server.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/orchestration/__init__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/orchestration/agent_router.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/orchestration/task_orchestrator.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/profiles/__init__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/profiles/base.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/profiles/default.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/quality_gates/__init__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/quality_gates/coordinator.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/quality_gates/diff_validator.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/quality_gates/evidence.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/quality_gates/failure_handler.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/quality_gates/functional_verifier.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/quality_gates/preflight_checks.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/quality_gates/success_criteria.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/quality_gates/test_validator.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/quality_gates/truth_enforcer.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/storage/task_type_manager.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/storage/work_queue.py +0 -0
- {sugarai-3.1.1.dev4/sugar/storage → sugarai-3.3.1/sugar/utils}/__init__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/utils/git_operations.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/utils/toon_encoder.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/workflow/__init__.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugar/workflow/orchestrator.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugarai.egg-info/dependency_links.txt +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugarai.egg-info/entry_points.txt +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/sugarai.egg-info/top_level.txt +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_billing.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_hold_functionality.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_hooks.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_orchestration.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_profiles.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_quality_gates.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_storage.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_task_types.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_toon_encoder.py +0 -0
- {sugarai-3.1.1.dev4 → sugarai-3.3.1}/tests/test_v3_benchmarks.py +0 -0
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: sugarai
|
|
3
|
-
Version: 3.
|
|
4
|
-
Summary: Sugar 🍰 -
|
|
3
|
+
Version: 3.3.1
|
|
4
|
+
Summary: Sugar 🍰 - Vibe coding while you sleep. Autonomous AI development for Claude Code.
|
|
5
5
|
Author-email: Steven Leggett <contact@roboticforce.io>
|
|
6
6
|
License: MIT
|
|
7
7
|
Project-URL: Homepage, https://github.com/roboticforce/sugar
|
|
@@ -50,13 +50,17 @@ Provides-Extra: test
|
|
|
50
50
|
Requires-Dist: pytest>=7.0.0; extra == "test"
|
|
51
51
|
Requires-Dist: pytest-asyncio>=0.21.0; extra == "test"
|
|
52
52
|
Requires-Dist: pytest-cov>=4.0.0; extra == "test"
|
|
53
|
+
Provides-Extra: mcp
|
|
54
|
+
Requires-Dist: mcp>=1.0.0; extra == "mcp"
|
|
55
|
+
Requires-Dist: starlette>=0.27.0; extra == "mcp"
|
|
56
|
+
Requires-Dist: uvicorn>=0.22.0; extra == "mcp"
|
|
53
57
|
Dynamic: license-file
|
|
54
58
|
|
|
55
59
|
# Sugar 🍰
|
|
56
60
|
|
|
57
|
-
|
|
61
|
+
Vibe coding while you sleep.
|
|
58
62
|
|
|
59
|
-
|
|
63
|
+
Autonomous AI development for Claude Code. Sugar builds features, fixes bugs, and ships code while you focus on what matters.
|
|
60
64
|
|
|
61
65
|
## What It Does
|
|
62
66
|
|
|
@@ -208,6 +212,108 @@ sugar list --status completed
|
|
|
208
212
|
- Confidence-based auto-posting
|
|
209
213
|
- Searchable issue history
|
|
210
214
|
|
|
215
|
+
**Ralph Wiggum Integration** *(New)*
|
|
216
|
+
- Iterative execution for complex tasks
|
|
217
|
+
- Self-correcting loops until tests pass
|
|
218
|
+
- Prevents single-shot failures
|
|
219
|
+
- Automatic completion detection
|
|
220
|
+
|
|
221
|
+
## Ralph Wiggum: Why Sugar Gets It Right
|
|
222
|
+
|
|
223
|
+
Here's the thing about AI coding: **single-shot attempts often fail on complex tasks**.
|
|
224
|
+
|
|
225
|
+
Ask Claude to implement a feature in one go, and you might get something that's 80% right. But that 20% means broken tests, edge cases missed, or subtle bugs. You end up going back and forth, manually iterating until it works.
|
|
226
|
+
|
|
227
|
+
**Ralph Wiggum fixes this by design.**
|
|
228
|
+
|
|
229
|
+
Instead of trying to complete a task perfectly the first time, Sugar can feed the same prompt repeatedly. Each iteration:
|
|
230
|
+
|
|
231
|
+
1. Claude sees its previous work in the files
|
|
232
|
+
2. Runs tests and sees what's failing
|
|
233
|
+
3. Fixes issues and improves the implementation
|
|
234
|
+
4. Repeats until the task is actually complete
|
|
235
|
+
|
|
236
|
+
```bash
|
|
237
|
+
# Without Ralph (traditional single-shot):
|
|
238
|
+
sugar add "Implement rate limiting"
|
|
239
|
+
# Claude attempts once, maybe tests fail, task marked "done" anyway
|
|
240
|
+
|
|
241
|
+
# With Ralph (iterative):
|
|
242
|
+
sugar add "Implement rate limiting" --ralph --max-iterations 10
|
|
243
|
+
# Claude iterates: implement → test → fix → test → fix → done
|
|
244
|
+
# Only marked complete when tests actually pass
|
|
245
|
+
```
|
|
246
|
+
|
|
247
|
+
**Think of it like code review cycles**, but automated. Junior dev writes code, tests fail, they fix it, tests pass, PR merged. Ralph does this loop automatically.
|
|
248
|
+
|
|
249
|
+
### When to Use Ralph
|
|
250
|
+
|
|
251
|
+
| Task Type | Without Ralph | With Ralph |
|
|
252
|
+
|-----------|---------------|------------|
|
|
253
|
+
| Simple bug fix | Works fine | Overkill |
|
|
254
|
+
| New feature | Hit or miss | Iterates until working |
|
|
255
|
+
| Complex refactor | Often breaks things | Self-corrects |
|
|
256
|
+
| TDD implementation | Tests often fail | Keeps going until green |
|
|
257
|
+
| Flaky test debugging | Might give up | Tries multiple approaches |
|
|
258
|
+
|
|
259
|
+
### How It Works
|
|
260
|
+
|
|
261
|
+
```
|
|
262
|
+
Iteration 1: "Implement rate limiting"
|
|
263
|
+
→ Creates RateLimiter class
|
|
264
|
+
→ Tests: 2 passing, 3 failing
|
|
265
|
+
|
|
266
|
+
Iteration 2: Same prompt, sees previous work
|
|
267
|
+
→ Fixes failing tests
|
|
268
|
+
→ Tests: 4 passing, 1 failing
|
|
269
|
+
|
|
270
|
+
Iteration 3: Same prompt, sees more progress
|
|
271
|
+
→ Handles edge case
|
|
272
|
+
→ Tests: 5 passing, 0 failing
|
|
273
|
+
→ Outputs: <promise>DONE</promise>
|
|
274
|
+
→ Task complete!
|
|
275
|
+
```
|
|
276
|
+
|
|
277
|
+
The `<promise>` tag is how Claude signals "I'm actually done." Without it, Ralph knows to keep iterating.
|
|
278
|
+
|
|
279
|
+
### Setup
|
|
280
|
+
|
|
281
|
+
**Ralph is built into Sugar** - no separate installation required. Just enable it in your config:
|
|
282
|
+
|
|
283
|
+
```yaml
|
|
284
|
+
# .sugar/config.yaml
|
|
285
|
+
sugar:
|
|
286
|
+
ralph:
|
|
287
|
+
enabled: true
|
|
288
|
+
max_iterations: 10
|
|
289
|
+
require_completion_criteria: true
|
|
290
|
+
```
|
|
291
|
+
|
|
292
|
+
Or use the `--ralph` flag when adding tasks:
|
|
293
|
+
|
|
294
|
+
```bash
|
|
295
|
+
sugar add "Complex feature" --ralph --max-iterations 15
|
|
296
|
+
```
|
|
297
|
+
|
|
298
|
+
### Safety First
|
|
299
|
+
|
|
300
|
+
Ralph won't run forever. You must include:
|
|
301
|
+
- A `<promise>` tag in your prompt (completion signal)
|
|
302
|
+
- OR `--max-iterations` flag (safety limit)
|
|
303
|
+
|
|
304
|
+
Sugar validates this BEFORE starting. No completion criteria = task rejected.
|
|
305
|
+
|
|
306
|
+
### Interactive Use (Claude Code)
|
|
307
|
+
|
|
308
|
+
For interactive Ralph loops in Claude Code sessions (outside of Sugar), install the Ralph Wiggum plugin:
|
|
309
|
+
|
|
310
|
+
```bash
|
|
311
|
+
# If you have the toolkit installed:
|
|
312
|
+
/ralph-wiggum:ralph-loop "Fix the flaky tests" --max-iterations 10
|
|
313
|
+
```
|
|
314
|
+
|
|
315
|
+
**Full docs:** [docs/ralph-wiggum.md](docs/ralph-wiggum.md)
|
|
316
|
+
|
|
211
317
|
## Issue Responder
|
|
212
318
|
|
|
213
319
|
Automatically analyze and respond to GitHub issues with AI-powered insights. Sugar understands issue context, codebase structure, and project patterns to generate helpful responses.
|
|
@@ -228,6 +334,16 @@ sugar issue respond 42 --post
|
|
|
228
334
|
|
|
229
335
|
The Issue Responder evaluates confidence before posting. Use `--force-post` to override the confidence check, or adjust the threshold with `--confidence-threshold`.
|
|
230
336
|
|
|
337
|
+
**Custom Prompts:** Customize Sugar's responses per-project by creating `.sugar/prompts/issue_responder.json`:
|
|
338
|
+
|
|
339
|
+
```json
|
|
340
|
+
{
|
|
341
|
+
"instructions": "You are a helpful assistant for MyProject. Be friendly and professional.",
|
|
342
|
+
"guidelines": ["Always search the codebase first", "Include file paths"],
|
|
343
|
+
"constraints": ["Never share API keys", "Don't promise release dates"]
|
|
344
|
+
}
|
|
345
|
+
```
|
|
346
|
+
|
|
231
347
|
**Full documentation:** [docs/issue-responder.md](docs/issue-responder.md)
|
|
232
348
|
|
|
233
349
|
## How It Works
|
|
@@ -1,8 +1,8 @@
|
|
|
1
1
|
# Sugar 🍰
|
|
2
2
|
|
|
3
|
-
|
|
3
|
+
Vibe coding while you sleep.
|
|
4
4
|
|
|
5
|
-
|
|
5
|
+
Autonomous AI development for Claude Code. Sugar builds features, fixes bugs, and ships code while you focus on what matters.
|
|
6
6
|
|
|
7
7
|
## What It Does
|
|
8
8
|
|
|
@@ -154,6 +154,108 @@ sugar list --status completed
|
|
|
154
154
|
- Confidence-based auto-posting
|
|
155
155
|
- Searchable issue history
|
|
156
156
|
|
|
157
|
+
**Ralph Wiggum Integration** *(New)*
|
|
158
|
+
- Iterative execution for complex tasks
|
|
159
|
+
- Self-correcting loops until tests pass
|
|
160
|
+
- Prevents single-shot failures
|
|
161
|
+
- Automatic completion detection
|
|
162
|
+
|
|
163
|
+
## Ralph Wiggum: Why Sugar Gets It Right
|
|
164
|
+
|
|
165
|
+
Here's the thing about AI coding: **single-shot attempts often fail on complex tasks**.
|
|
166
|
+
|
|
167
|
+
Ask Claude to implement a feature in one go, and you might get something that's 80% right. But that 20% means broken tests, edge cases missed, or subtle bugs. You end up going back and forth, manually iterating until it works.
|
|
168
|
+
|
|
169
|
+
**Ralph Wiggum fixes this by design.**
|
|
170
|
+
|
|
171
|
+
Instead of trying to complete a task perfectly the first time, Sugar can feed the same prompt repeatedly. Each iteration:
|
|
172
|
+
|
|
173
|
+
1. Claude sees its previous work in the files
|
|
174
|
+
2. Runs tests and sees what's failing
|
|
175
|
+
3. Fixes issues and improves the implementation
|
|
176
|
+
4. Repeats until the task is actually complete
|
|
177
|
+
|
|
178
|
+
```bash
|
|
179
|
+
# Without Ralph (traditional single-shot):
|
|
180
|
+
sugar add "Implement rate limiting"
|
|
181
|
+
# Claude attempts once, maybe tests fail, task marked "done" anyway
|
|
182
|
+
|
|
183
|
+
# With Ralph (iterative):
|
|
184
|
+
sugar add "Implement rate limiting" --ralph --max-iterations 10
|
|
185
|
+
# Claude iterates: implement → test → fix → test → fix → done
|
|
186
|
+
# Only marked complete when tests actually pass
|
|
187
|
+
```
|
|
188
|
+
|
|
189
|
+
**Think of it like code review cycles**, but automated. Junior dev writes code, tests fail, they fix it, tests pass, PR merged. Ralph does this loop automatically.
|
|
190
|
+
|
|
191
|
+
### When to Use Ralph
|
|
192
|
+
|
|
193
|
+
| Task Type | Without Ralph | With Ralph |
|
|
194
|
+
|-----------|---------------|------------|
|
|
195
|
+
| Simple bug fix | Works fine | Overkill |
|
|
196
|
+
| New feature | Hit or miss | Iterates until working |
|
|
197
|
+
| Complex refactor | Often breaks things | Self-corrects |
|
|
198
|
+
| TDD implementation | Tests often fail | Keeps going until green |
|
|
199
|
+
| Flaky test debugging | Might give up | Tries multiple approaches |
|
|
200
|
+
|
|
201
|
+
### How It Works
|
|
202
|
+
|
|
203
|
+
```
|
|
204
|
+
Iteration 1: "Implement rate limiting"
|
|
205
|
+
→ Creates RateLimiter class
|
|
206
|
+
→ Tests: 2 passing, 3 failing
|
|
207
|
+
|
|
208
|
+
Iteration 2: Same prompt, sees previous work
|
|
209
|
+
→ Fixes failing tests
|
|
210
|
+
→ Tests: 4 passing, 1 failing
|
|
211
|
+
|
|
212
|
+
Iteration 3: Same prompt, sees more progress
|
|
213
|
+
→ Handles edge case
|
|
214
|
+
→ Tests: 5 passing, 0 failing
|
|
215
|
+
→ Outputs: <promise>DONE</promise>
|
|
216
|
+
→ Task complete!
|
|
217
|
+
```
|
|
218
|
+
|
|
219
|
+
The `<promise>` tag is how Claude signals "I'm actually done." Without it, Ralph knows to keep iterating.
|
|
220
|
+
|
|
221
|
+
### Setup
|
|
222
|
+
|
|
223
|
+
**Ralph is built into Sugar** - no separate installation required. Just enable it in your config:
|
|
224
|
+
|
|
225
|
+
```yaml
|
|
226
|
+
# .sugar/config.yaml
|
|
227
|
+
sugar:
|
|
228
|
+
ralph:
|
|
229
|
+
enabled: true
|
|
230
|
+
max_iterations: 10
|
|
231
|
+
require_completion_criteria: true
|
|
232
|
+
```
|
|
233
|
+
|
|
234
|
+
Or use the `--ralph` flag when adding tasks:
|
|
235
|
+
|
|
236
|
+
```bash
|
|
237
|
+
sugar add "Complex feature" --ralph --max-iterations 15
|
|
238
|
+
```
|
|
239
|
+
|
|
240
|
+
### Safety First
|
|
241
|
+
|
|
242
|
+
Ralph won't run forever. You must include:
|
|
243
|
+
- A `<promise>` tag in your prompt (completion signal)
|
|
244
|
+
- OR `--max-iterations` flag (safety limit)
|
|
245
|
+
|
|
246
|
+
Sugar validates this BEFORE starting. No completion criteria = task rejected.
|
|
247
|
+
|
|
248
|
+
### Interactive Use (Claude Code)
|
|
249
|
+
|
|
250
|
+
For interactive Ralph loops in Claude Code sessions (outside of Sugar), install the Ralph Wiggum plugin:
|
|
251
|
+
|
|
252
|
+
```bash
|
|
253
|
+
# If you have the toolkit installed:
|
|
254
|
+
/ralph-wiggum:ralph-loop "Fix the flaky tests" --max-iterations 10
|
|
255
|
+
```
|
|
256
|
+
|
|
257
|
+
**Full docs:** [docs/ralph-wiggum.md](docs/ralph-wiggum.md)
|
|
258
|
+
|
|
157
259
|
## Issue Responder
|
|
158
260
|
|
|
159
261
|
Automatically analyze and respond to GitHub issues with AI-powered insights. Sugar understands issue context, codebase structure, and project patterns to generate helpful responses.
|
|
@@ -174,6 +276,16 @@ sugar issue respond 42 --post
|
|
|
174
276
|
|
|
175
277
|
The Issue Responder evaluates confidence before posting. Use `--force-post` to override the confidence check, or adjust the threshold with `--confidence-threshold`.
|
|
176
278
|
|
|
279
|
+
**Custom Prompts:** Customize Sugar's responses per-project by creating `.sugar/prompts/issue_responder.json`:
|
|
280
|
+
|
|
281
|
+
```json
|
|
282
|
+
{
|
|
283
|
+
"instructions": "You are a helpful assistant for MyProject. Be friendly and professional.",
|
|
284
|
+
"guidelines": ["Always search the codebase first", "Include file paths"],
|
|
285
|
+
"constraints": ["Never share API keys", "Don't promise release dates"]
|
|
286
|
+
}
|
|
287
|
+
```
|
|
288
|
+
|
|
177
289
|
**Full documentation:** [docs/issue-responder.md](docs/issue-responder.md)
|
|
178
290
|
|
|
179
291
|
## How It Works
|
|
@@ -4,8 +4,8 @@ build-backend = "setuptools.build_meta"
|
|
|
4
4
|
|
|
5
5
|
[project]
|
|
6
6
|
name = "sugarai"
|
|
7
|
-
version = "3.
|
|
8
|
-
description = "Sugar 🍰 -
|
|
7
|
+
version = "3.3.1"
|
|
8
|
+
description = "Sugar 🍰 - Vibe coding while you sleep. Autonomous AI development for Claude Code."
|
|
9
9
|
readme = "README.md"
|
|
10
10
|
requires-python = ">=3.11"
|
|
11
11
|
license = {text = "MIT"}
|
|
@@ -59,6 +59,11 @@ test = [
|
|
|
59
59
|
"pytest-asyncio>=0.21.0",
|
|
60
60
|
"pytest-cov>=4.0.0"
|
|
61
61
|
]
|
|
62
|
+
mcp = [
|
|
63
|
+
"mcp>=1.0.0",
|
|
64
|
+
"starlette>=0.27.0",
|
|
65
|
+
"uvicorn>=0.22.0",
|
|
66
|
+
]
|
|
62
67
|
|
|
63
68
|
[project.scripts]
|
|
64
69
|
sugar = "sugar.main:cli"
|
|
@@ -134,8 +134,8 @@ class SugarAgentConfig:
|
|
|
134
134
|
# Working directory
|
|
135
135
|
working_directory: Optional[str] = None
|
|
136
136
|
|
|
137
|
-
# Timeout for operations (seconds)
|
|
138
|
-
timeout: int =
|
|
137
|
+
# Timeout for operations (seconds) - default 2 hours for complex tasks
|
|
138
|
+
timeout: int = 7200
|
|
139
139
|
|
|
140
140
|
# Retry settings for transient errors
|
|
141
141
|
max_retries: int = 3
|
|
@@ -299,60 +299,76 @@ Guidelines:
|
|
|
299
299
|
files_modified = []
|
|
300
300
|
|
|
301
301
|
# Use the SDK's query() function which returns an async generator
|
|
302
|
-
|
|
303
|
-
|
|
304
|
-
|
|
305
|
-
|
|
306
|
-
|
|
307
|
-
|
|
308
|
-
if isinstance(
|
|
309
|
-
|
|
310
|
-
|
|
311
|
-
|
|
312
|
-
|
|
313
|
-
|
|
314
|
-
|
|
315
|
-
|
|
316
|
-
|
|
317
|
-
|
|
318
|
-
|
|
319
|
-
|
|
320
|
-
|
|
321
|
-
|
|
322
|
-
|
|
323
|
-
|
|
324
|
-
|
|
325
|
-
|
|
326
|
-
|
|
327
|
-
|
|
328
|
-
|
|
329
|
-
|
|
330
|
-
|
|
331
|
-
|
|
332
|
-
|
|
333
|
-
|
|
334
|
-
|
|
335
|
-
|
|
336
|
-
|
|
337
|
-
"
|
|
338
|
-
|
|
339
|
-
|
|
340
|
-
|
|
341
|
-
|
|
342
|
-
|
|
343
|
-
|
|
344
|
-
|
|
345
|
-
|
|
346
|
-
|
|
347
|
-
|
|
348
|
-
|
|
349
|
-
|
|
350
|
-
|
|
351
|
-
|
|
352
|
-
|
|
353
|
-
|
|
354
|
-
|
|
355
|
-
|
|
302
|
+
# Wrap with timeout to prevent hanging if generator doesn't signal completion
|
|
303
|
+
try:
|
|
304
|
+
async with asyncio.timeout(self.config.timeout):
|
|
305
|
+
async for message in query(prompt=prompt, options=options):
|
|
306
|
+
# Handle different message types from the SDK
|
|
307
|
+
# The SDK may return dicts or typed objects depending on version
|
|
308
|
+
if SDK_HAS_TYPES and isinstance(message, AssistantMessage):
|
|
309
|
+
# Typed SDK - iterate through content blocks
|
|
310
|
+
for block in message.content:
|
|
311
|
+
if isinstance(block, TextBlock):
|
|
312
|
+
content_parts.append(block.text)
|
|
313
|
+
elif isinstance(block, ToolUseBlock):
|
|
314
|
+
tool_use = {
|
|
315
|
+
"tool": block.name,
|
|
316
|
+
"input": block.input,
|
|
317
|
+
}
|
|
318
|
+
tool_uses.append(tool_use)
|
|
319
|
+
|
|
320
|
+
# Track file modifications
|
|
321
|
+
if block.name in ("Write", "Edit"):
|
|
322
|
+
file_path = block.input.get("file_path")
|
|
323
|
+
if file_path and file_path not in files_modified:
|
|
324
|
+
files_modified.append(file_path)
|
|
325
|
+
|
|
326
|
+
elif isinstance(message, dict):
|
|
327
|
+
# Dict-based SDK response
|
|
328
|
+
msg_type = message.get("type", "")
|
|
329
|
+
|
|
330
|
+
if msg_type == "assistant":
|
|
331
|
+
# Process content blocks
|
|
332
|
+
content = message.get("content", [])
|
|
333
|
+
for block in content:
|
|
334
|
+
block_type = block.get("type", "")
|
|
335
|
+
if block_type == "text":
|
|
336
|
+
content_parts.append(block.get("text", ""))
|
|
337
|
+
elif block_type == "tool_use":
|
|
338
|
+
tool_use = {
|
|
339
|
+
"tool": block.get("name", ""),
|
|
340
|
+
"input": block.get("input", {}),
|
|
341
|
+
}
|
|
342
|
+
tool_uses.append(tool_use)
|
|
343
|
+
|
|
344
|
+
# Track file modifications
|
|
345
|
+
tool_name = block.get("name", "")
|
|
346
|
+
if tool_name in ("Write", "Edit"):
|
|
347
|
+
file_path = block.get("input", {}).get(
|
|
348
|
+
"file_path"
|
|
349
|
+
)
|
|
350
|
+
if (
|
|
351
|
+
file_path
|
|
352
|
+
and file_path not in files_modified
|
|
353
|
+
):
|
|
354
|
+
files_modified.append(file_path)
|
|
355
|
+
|
|
356
|
+
elif msg_type == "text":
|
|
357
|
+
# Direct text message
|
|
358
|
+
content_parts.append(message.get("text", ""))
|
|
359
|
+
|
|
360
|
+
elif msg_type == "result":
|
|
361
|
+
# Final result message
|
|
362
|
+
if message.get("content"):
|
|
363
|
+
content_parts.append(str(message.get("content", "")))
|
|
364
|
+
|
|
365
|
+
except asyncio.TimeoutError:
|
|
366
|
+
logger.warning(
|
|
367
|
+
f"Query timed out after {self.config.timeout}s. "
|
|
368
|
+
f"Returning partial results: {len(tool_uses)} tool uses, "
|
|
369
|
+
f"{len(files_modified)} files modified"
|
|
370
|
+
)
|
|
371
|
+
# Return partial results - the work done so far is still valid
|
|
356
372
|
|
|
357
373
|
return content_parts, tool_uses, files_modified
|
|
358
374
|
|
|
@@ -0,0 +1,143 @@
|
|
|
1
|
+
"""Configuration loader for the Issue Responder feature."""
|
|
2
|
+
|
|
3
|
+
from dataclasses import dataclass, field
|
|
4
|
+
from typing import List, Optional
|
|
5
|
+
import yaml
|
|
6
|
+
from pathlib import Path
|
|
7
|
+
|
|
8
|
+
|
|
9
|
+
@dataclass
|
|
10
|
+
class IssueResponderConfig:
|
|
11
|
+
"""Configuration for AI-powered automatic issue responses.
|
|
12
|
+
|
|
13
|
+
Attributes:
|
|
14
|
+
enabled: Enable automatic issue responses
|
|
15
|
+
auto_post_threshold: Minimum confidence score (0.0-1.0) to auto-post responses
|
|
16
|
+
max_response_length: Maximum characters per response
|
|
17
|
+
response_delay_seconds: Delay before posting (0 = immediate)
|
|
18
|
+
rate_limit_per_hour: Maximum responses per hour (0 = unlimited)
|
|
19
|
+
respond_to_labels: Only respond to issues with these labels (empty = all)
|
|
20
|
+
skip_labels: Never respond to issues with these labels
|
|
21
|
+
skip_bot_issues: Skip issues created by bots
|
|
22
|
+
handle_follow_ups: How to handle follow-up comments - "ignore" | "queue" | "auto"
|
|
23
|
+
model: Claude model to use for generating responses
|
|
24
|
+
"""
|
|
25
|
+
|
|
26
|
+
enabled: bool = False
|
|
27
|
+
auto_post_threshold: float = 0.8
|
|
28
|
+
max_response_length: int = 2000
|
|
29
|
+
response_delay_seconds: int = 0
|
|
30
|
+
rate_limit_per_hour: int = 10
|
|
31
|
+
respond_to_labels: List[str] = field(default_factory=list)
|
|
32
|
+
skip_labels: List[str] = field(
|
|
33
|
+
default_factory=lambda: ["wontfix", "duplicate", "stale"]
|
|
34
|
+
)
|
|
35
|
+
skip_bot_issues: bool = True
|
|
36
|
+
handle_follow_ups: str = "ignore" # "ignore" | "queue" | "auto"
|
|
37
|
+
model: str = "claude-sonnet-4-20250514"
|
|
38
|
+
|
|
39
|
+
@classmethod
|
|
40
|
+
def from_dict(cls, data: dict) -> "IssueResponderConfig":
|
|
41
|
+
"""Create IssueResponderConfig from a dictionary.
|
|
42
|
+
|
|
43
|
+
Args:
|
|
44
|
+
data: Dictionary containing configuration values
|
|
45
|
+
|
|
46
|
+
Returns:
|
|
47
|
+
IssueResponderConfig instance
|
|
48
|
+
|
|
49
|
+
Example:
|
|
50
|
+
>>> config_data = {
|
|
51
|
+
... "enabled": True,
|
|
52
|
+
... "auto_post_threshold": 0.9,
|
|
53
|
+
... "respond_to_labels": ["question", "help-wanted"]
|
|
54
|
+
... }
|
|
55
|
+
>>> config = IssueResponderConfig.from_dict(config_data)
|
|
56
|
+
"""
|
|
57
|
+
# Filter to only include known fields
|
|
58
|
+
valid_fields = {f.name for f in cls.__dataclass_fields__.values()}
|
|
59
|
+
filtered_data = {k: v for k, v in data.items() if k in valid_fields}
|
|
60
|
+
return cls(**filtered_data)
|
|
61
|
+
|
|
62
|
+
@classmethod
|
|
63
|
+
def load_from_file(cls, config_path: str) -> "IssueResponderConfig":
|
|
64
|
+
"""Load IssueResponderConfig from a YAML configuration file.
|
|
65
|
+
|
|
66
|
+
Args:
|
|
67
|
+
config_path: Path to the YAML configuration file
|
|
68
|
+
|
|
69
|
+
Returns:
|
|
70
|
+
IssueResponderConfig instance with values from the file
|
|
71
|
+
|
|
72
|
+
Raises:
|
|
73
|
+
FileNotFoundError: If the config file doesn't exist
|
|
74
|
+
KeyError: If the 'issue_responder' section is missing
|
|
75
|
+
yaml.YAMLError: If the YAML file is malformed
|
|
76
|
+
|
|
77
|
+
Example:
|
|
78
|
+
>>> config = IssueResponderConfig.load_from_file(".sugar/config.yaml")
|
|
79
|
+
"""
|
|
80
|
+
config_file = Path(config_path)
|
|
81
|
+
|
|
82
|
+
if not config_file.exists():
|
|
83
|
+
raise FileNotFoundError(f"Configuration file not found: {config_path}")
|
|
84
|
+
|
|
85
|
+
with open(config_file, "r") as f:
|
|
86
|
+
config_data = yaml.safe_load(f)
|
|
87
|
+
|
|
88
|
+
# Navigate to the issue_responder section
|
|
89
|
+
if "sugar" not in config_data:
|
|
90
|
+
raise KeyError("Configuration file missing 'sugar' section")
|
|
91
|
+
|
|
92
|
+
if "issue_responder" not in config_data["sugar"]:
|
|
93
|
+
raise KeyError("Configuration file missing 'sugar.issue_responder' section")
|
|
94
|
+
|
|
95
|
+
issue_responder_data = config_data["sugar"]["issue_responder"]
|
|
96
|
+
return cls.from_dict(issue_responder_data)
|
|
97
|
+
|
|
98
|
+
def validate(self) -> List[str]:
|
|
99
|
+
"""Validate configuration values and return list of errors.
|
|
100
|
+
|
|
101
|
+
Returns:
|
|
102
|
+
List of validation error messages (empty if valid)
|
|
103
|
+
|
|
104
|
+
Example:
|
|
105
|
+
>>> config = IssueResponderConfig(auto_post_threshold=1.5)
|
|
106
|
+
>>> errors = config.validate()
|
|
107
|
+
>>> if errors:
|
|
108
|
+
... print("Invalid configuration:", errors)
|
|
109
|
+
"""
|
|
110
|
+
errors = []
|
|
111
|
+
|
|
112
|
+
# Validate auto_post_threshold
|
|
113
|
+
if not 0.0 <= self.auto_post_threshold <= 1.0:
|
|
114
|
+
errors.append(
|
|
115
|
+
f"auto_post_threshold must be between 0.0 and 1.0, got {self.auto_post_threshold}"
|
|
116
|
+
)
|
|
117
|
+
|
|
118
|
+
# Validate max_response_length
|
|
119
|
+
if self.max_response_length <= 0:
|
|
120
|
+
errors.append(
|
|
121
|
+
f"max_response_length must be positive, got {self.max_response_length}"
|
|
122
|
+
)
|
|
123
|
+
|
|
124
|
+
# Validate response_delay_seconds
|
|
125
|
+
if self.response_delay_seconds < 0:
|
|
126
|
+
errors.append(
|
|
127
|
+
f"response_delay_seconds must be non-negative, got {self.response_delay_seconds}"
|
|
128
|
+
)
|
|
129
|
+
|
|
130
|
+
# Validate rate_limit_per_hour
|
|
131
|
+
if self.rate_limit_per_hour < 0:
|
|
132
|
+
errors.append(
|
|
133
|
+
f"rate_limit_per_hour must be non-negative, got {self.rate_limit_per_hour}"
|
|
134
|
+
)
|
|
135
|
+
|
|
136
|
+
# Validate handle_follow_ups
|
|
137
|
+
valid_follow_up_values = {"ignore", "queue", "auto"}
|
|
138
|
+
if self.handle_follow_ups not in valid_follow_up_values:
|
|
139
|
+
errors.append(
|
|
140
|
+
f"handle_follow_ups must be one of {valid_follow_up_values}, got '{self.handle_follow_ups}'"
|
|
141
|
+
)
|
|
142
|
+
|
|
143
|
+
return errors
|