@covibes/zeroshot 5.4.0 → 5.4.1
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/LICENSE +1 -1
- package/README.md +8 -465
- package/cli/index.js +101 -5231
- package/package.json +16 -130
- package/CHANGELOG.md +0 -537
- package/cli/commands/providers.js +0 -149
- package/cli/lib/first-run.js +0 -211
- package/cli/lib/update-checker.js +0 -281
- package/cli/message-formatter-utils.js +0 -75
- package/cli/message-formatters-normal.js +0 -275
- package/cli/message-formatters-watch.js +0 -181
- package/cluster-templates/base-templates/debug-workflow.json +0 -375
- package/cluster-templates/base-templates/full-workflow.json +0 -581
- package/cluster-templates/base-templates/single-worker.json +0 -63
- package/cluster-templates/base-templates/worker-validator.json +0 -208
- package/cluster-templates/conductor-bootstrap.json +0 -124
- package/docker/zeroshot-cluster/Dockerfile +0 -152
- package/docker/zeroshot-cluster/pre-baked-deps.json +0 -28
- package/lib/completion.js +0 -174
- package/lib/docker-config.js +0 -220
- package/lib/git-remote-utils.js +0 -165
- package/lib/id-detector.js +0 -56
- package/lib/provider-defaults.js +0 -62
- package/lib/provider-detection.js +0 -59
- package/lib/provider-names.js +0 -57
- package/lib/settings/claude-auth.js +0 -78
- package/lib/settings.js +0 -503
- package/lib/stream-json-parser.js +0 -10
- package/scripts/fix-node-pty-permissions.js +0 -75
- package/scripts/record-demo.sh +0 -279
- package/scripts/setup-merge-queue.sh +0 -170
- package/scripts/test-install.sh +0 -40
- package/scripts/validate-templates.js +0 -100
- package/src/agent/agent-config.js +0 -246
- package/src/agent/agent-context-builder.js +0 -444
- package/src/agent/agent-hook-executor.js +0 -599
- package/src/agent/agent-lifecycle.js +0 -743
- package/src/agent/agent-stuck-detector.js +0 -256
- package/src/agent/agent-task-executor.js +0 -1690
- package/src/agent/agent-trigger-evaluator.js +0 -67
- package/src/agent/output-extraction.js +0 -217
- package/src/agent/output-reformatter.js +0 -175
- package/src/agent/schema-utils.js +0 -146
- package/src/agent-wrapper.js +0 -579
- package/src/agents/git-pusher-template.js +0 -285
- package/src/attach/attach-client.js +0 -438
- package/src/attach/attach-server.js +0 -543
- package/src/attach/index.js +0 -35
- package/src/attach/protocol.js +0 -220
- package/src/attach/ring-buffer.js +0 -121
- package/src/attach/socket-discovery.js +0 -267
- package/src/claude-task-runner.js +0 -574
- package/src/config-router.js +0 -80
- package/src/config-validator.js +0 -1925
- package/src/copy-worker.js +0 -43
- package/src/input-helpers.js +0 -65
- package/src/isolation-manager.js +0 -1471
- package/src/issue-providers/README.md +0 -305
- package/src/issue-providers/azure-devops-provider.js +0 -273
- package/src/issue-providers/base-provider.js +0 -232
- package/src/issue-providers/github-provider.js +0 -179
- package/src/issue-providers/gitlab-provider.js +0 -241
- package/src/issue-providers/index.js +0 -196
- package/src/issue-providers/jira-provider.js +0 -239
- package/src/ledger.js +0 -631
- package/src/lib/safe-exec.js +0 -88
- package/src/logic-engine.js +0 -223
- package/src/message-bus-bridge.js +0 -139
- package/src/message-bus.js +0 -249
- package/src/name-generator.js +0 -232
- package/src/orchestrator.js +0 -2815
- package/src/preflight.js +0 -639
- package/src/process-metrics.js +0 -608
- package/src/providers/anthropic/cli-builder.js +0 -73
- package/src/providers/anthropic/index.js +0 -204
- package/src/providers/anthropic/models.js +0 -23
- package/src/providers/anthropic/output-parser.js +0 -177
- package/src/providers/base-provider.js +0 -251
- package/src/providers/capabilities.js +0 -60
- package/src/providers/google/cli-builder.js +0 -55
- package/src/providers/google/index.js +0 -116
- package/src/providers/google/models.js +0 -24
- package/src/providers/google/output-parser.js +0 -101
- package/src/providers/index.js +0 -91
- package/src/providers/openai/cli-builder.js +0 -133
- package/src/providers/openai/index.js +0 -136
- package/src/providers/openai/models.js +0 -21
- package/src/providers/openai/output-parser.js +0 -143
- package/src/providers/opencode/cli-builder.js +0 -42
- package/src/providers/opencode/index.js +0 -103
- package/src/providers/opencode/models.js +0 -55
- package/src/providers/opencode/output-parser.js +0 -122
- package/src/schemas/sub-cluster.js +0 -185
- package/src/status-footer.js +0 -1026
- package/src/sub-cluster-wrapper.js +0 -601
- package/src/task-runner.js +0 -30
- package/src/template-resolver.js +0 -372
- package/src/tui/CHANGES.txt +0 -133
- package/src/tui/LAYOUT.md +0 -261
- package/src/tui/README.txt +0 -192
- package/src/tui/TWO-LEVEL-NAVIGATION.md +0 -186
- package/src/tui/data-poller.js +0 -349
- package/src/tui/demo.js +0 -208
- package/src/tui/formatters.js +0 -129
- package/src/tui/index.js +0 -193
- package/src/tui/keybindings.js +0 -324
- package/src/tui/layout.js +0 -334
- package/src/tui/renderer.js +0 -184
- package/task-lib/attachable-watcher.js +0 -309
- package/task-lib/claude-recovery.js +0 -156
- package/task-lib/commands/clean.js +0 -50
- package/task-lib/commands/episodes.js +0 -105
- package/task-lib/commands/get-log-path.js +0 -23
- package/task-lib/commands/kill.js +0 -32
- package/task-lib/commands/list.js +0 -105
- package/task-lib/commands/logs.js +0 -453
- package/task-lib/commands/resume.js +0 -42
- package/task-lib/commands/run.js +0 -57
- package/task-lib/commands/schedule.js +0 -105
- package/task-lib/commands/scheduler-cmd.js +0 -96
- package/task-lib/commands/schedules.js +0 -148
- package/task-lib/commands/status.js +0 -44
- package/task-lib/commands/unschedule.js +0 -16
- package/task-lib/completion.js +0 -9
- package/task-lib/config.js +0 -8
- package/task-lib/name-generator.js +0 -230
- package/task-lib/package.json +0 -3
- package/task-lib/runner.js +0 -201
- package/task-lib/scheduler.js +0 -252
- package/task-lib/store.js +0 -529
- package/task-lib/tui/formatters.js +0 -215
- package/task-lib/tui/index.js +0 -197
- package/task-lib/tui/layout.js +0 -111
- package/task-lib/tui/renderer.js +0 -129
- package/task-lib/tui.js +0 -425
- package/task-lib/watcher.js +0 -226
package/LICENSE
CHANGED
package/README.md
CHANGED
|
@@ -1,475 +1,18 @@
|
|
|
1
|
-
# zeroshot
|
|
1
|
+
# @covibes/zeroshot
|
|
2
2
|
|
|
3
|
-
|
|
3
|
+
This package has moved to `@the-open-engine/zeroshot`.
|
|
4
4
|
|
|
5
|
-
|
|
6
|
-
<p align="center">
|
|
7
|
-
<code>npm install -g @covibes/zeroshot</code>
|
|
8
|
-
</p>
|
|
9
|
-
|
|
10
|
-
<p align="center">
|
|
11
|
-
<img src="./docs/assets/zeroshot-demo.gif" alt="Demo" width="700">
|
|
12
|
-
<br>
|
|
13
|
-
<em>Demo (100x speed, 90-minute run, 5 iterations to approval)</em>
|
|
14
|
-
</p>
|
|
15
|
-
|
|
16
|
-
[](https://github.com/covibes/zeroshot/actions/workflows/ci.yml)
|
|
17
|
-
[](https://www.npmjs.com/package/@covibes/zeroshot)
|
|
18
|
-
[](LICENSE)
|
|
19
|
-
[](https://nodejs.org/)
|
|
20
|
-

|
|
21
|
-
|
|
22
|
-
<!-- discord-placeholder -->
|
|
23
|
-
|
|
24
|
-
[](https://discord.gg/PdZ3UEXB)
|
|
25
|
-
|
|
26
|
-
Zeroshot is an open-source AI coding agent orchestration CLI that runs multi-agent workflows to autonomously implement, review, test, and verify code changes.
|
|
27
|
-
|
|
28
|
-
It runs a **planner**, an **implementer**, and independent **validators** in isolated environments, looping until changes are **verified** or **rejected** with actionable, reproducible failures.
|
|
29
|
-
|
|
30
|
-
Built for tasks where correctness matters more than speed.
|
|
31
|
-
|
|
32
|
-
## How It Works
|
|
33
|
-
|
|
34
|
-
- Plan: translate a task into concrete acceptance criteria
|
|
35
|
-
- Implement: make changes in an isolated workspace (local, worktree, or Docker)
|
|
36
|
-
- Validate: run automated checks with independent validators
|
|
37
|
-
- Iterate: repeat until verified, or return actionable failures
|
|
38
|
-
- Resume: crash-safe state persisted for recovery
|
|
39
|
-
|
|
40
|
-
## Quick Start
|
|
41
|
-
|
|
42
|
-
```bash
|
|
43
|
-
zeroshot run 123 # GitHub issue number
|
|
44
|
-
zeroshot run feature.md # Markdown file
|
|
45
|
-
zeroshot run "Add dark mode" # Inline text
|
|
46
|
-
```
|
|
47
|
-
|
|
48
|
-
Or describe a complex task inline:
|
|
49
|
-
|
|
50
|
-
```bash
|
|
51
|
-
zeroshot run "Add optimistic locking with automatic retry: when updating a user,
|
|
52
|
-
retry with exponential backoff up to 3 times, merge non-conflicting field changes,
|
|
53
|
-
and surface conflicts with details. Handle the ABA problem where version goes A->B->A."
|
|
54
|
-
```
|
|
55
|
-
|
|
56
|
-
## Why Not Just Use a Single AI Agent?
|
|
57
|
-
|
|
58
|
-
| Approach | Writes Code | Runs Tests | Blind Validation | Iterates Until Verified |
|
|
59
|
-
| -------------------------- | ----------- | ---------- | ---------------- | ----------------------- |
|
|
60
|
-
| Chat-based assistant | ✅ | ⚠️ | ❌ | ❌ |
|
|
61
|
-
| Single coding agent | ✅ | ⚠️ | ❌ | ⚠️ |
|
|
62
|
-
| **Zeroshot (multi-agent)** | ✅ | ✅ | ✅ | ✅ |
|
|
63
|
-
|
|
64
|
-
## Use Cases
|
|
65
|
-
|
|
66
|
-
- Autonomous AI code refactoring
|
|
67
|
-
- AI-powered pull request automation
|
|
68
|
-
- Automated bug fixing with validation
|
|
69
|
-
- Multi-agent code generation for software engineering
|
|
70
|
-
- Agentic coding workflows with blind validation
|
|
71
|
-
|
|
72
|
-
## Who Is This For?
|
|
73
|
-
|
|
74
|
-
- Senior engineers who care about correctness and reproducibility
|
|
75
|
-
- Teams automating PR workflows and code review gates
|
|
76
|
-
- Infra/platform teams standardizing agentic workflows
|
|
77
|
-
- Open-source maintainers working through issue backlogs
|
|
78
|
-
- AI power users who want verification, not vibes
|
|
79
|
-
|
|
80
|
-
## Install and Requirements
|
|
81
|
-
|
|
82
|
-
**Platforms**: Linux, macOS (Windows WSL not yet supported)
|
|
83
|
-
|
|
84
|
-
```bash
|
|
85
|
-
npm install -g @covibes/zeroshot
|
|
86
|
-
```
|
|
87
|
-
|
|
88
|
-
**Requires**: Node 18+, at least one provider CLI (Claude Code, Codex, Gemini, Opencode).
|
|
89
|
-
|
|
90
|
-
```bash
|
|
91
|
-
# Install one or more providers
|
|
92
|
-
npm i -g @anthropic-ai/claude-code
|
|
93
|
-
npm i -g @openai/codex
|
|
94
|
-
npm i -g @google/gemini-cli
|
|
95
|
-
# Opencode: see https://opencode.ai
|
|
96
|
-
|
|
97
|
-
# Authenticate with the provider CLI
|
|
98
|
-
claude login # Claude
|
|
99
|
-
codex login # Codex
|
|
100
|
-
gemini auth login # Gemini
|
|
101
|
-
opencode auth login # Opencode
|
|
102
|
-
|
|
103
|
-
# GitHub auth (for issue numbers)
|
|
104
|
-
gh auth login
|
|
105
|
-
```
|
|
106
|
-
|
|
107
|
-
## Providers
|
|
108
|
-
|
|
109
|
-
Zeroshot shells out to provider CLIs. Pick a default and override per run:
|
|
110
|
-
|
|
111
|
-
```bash
|
|
112
|
-
zeroshot providers
|
|
113
|
-
zeroshot providers set-default codex
|
|
114
|
-
zeroshot run 123 --provider gemini
|
|
115
|
-
```
|
|
116
|
-
|
|
117
|
-
See `docs/providers.md` for setup, model levels, and Docker mounts.
|
|
118
|
-
|
|
119
|
-
## Why Multiple Agents?
|
|
120
|
-
|
|
121
|
-
Single-agent sessions degrade. Context gets buried under thousands of tokens. The model optimizes for "done" over "correct."
|
|
122
|
-
|
|
123
|
-
Zeroshot fixes this with isolated agents that check each other's work. Validators can't lie about code they didn't write. Fail the check? Fix and retry until it actually works.
|
|
124
|
-
|
|
125
|
-
## What Makes It Different
|
|
126
|
-
|
|
127
|
-
- **Blind validation** - Validators never see the worker's context or code history
|
|
128
|
-
- **Repeatable workflows** - Task complexity determines agent count and model selection
|
|
129
|
-
- **Accept/reject loop** - Rejections include actionable findings, not vague complaints
|
|
130
|
-
- **Crash recovery** - All state persisted to SQLite; resume anytime
|
|
131
|
-
- **Isolation modes** - None, git worktree, or Docker container
|
|
132
|
-
- **Cost control** - Model ceilings prevent runaway API spend
|
|
133
|
-
|
|
134
|
-
## When to Use Zeroshot
|
|
135
|
-
|
|
136
|
-
Zeroshot performs best when tasks have clear acceptance criteria.
|
|
137
|
-
|
|
138
|
-
| Scenario | Use | Why |
|
|
139
|
-
| ----------------------------------------------- | --- | ------------------------- |
|
|
140
|
-
| Add rate limiting (sliding window, per-IP, 429) | Yes | Clear requirements |
|
|
141
|
-
| Refactor auth to JWT | Yes | Defined end state |
|
|
142
|
-
| Fix login bug | Yes | Success is measurable |
|
|
143
|
-
| Fix 2410 lint violations | Yes | Clear completion criteria |
|
|
144
|
-
| Make the app faster | No | Needs exploration first |
|
|
145
|
-
| Improve the codebase | No | No acceptance criteria |
|
|
146
|
-
| Figure out flaky tests | No | Exploratory |
|
|
147
|
-
|
|
148
|
-
Rule of thumb: if you cannot describe what "done" means, validators cannot verify it.
|
|
149
|
-
|
|
150
|
-
## Command Overview
|
|
151
|
-
|
|
152
|
-
```bash
|
|
153
|
-
# Run
|
|
154
|
-
zeroshot run 123 # GitHub issue
|
|
155
|
-
zeroshot run feature.md # Markdown file
|
|
156
|
-
zeroshot run "Add dark mode" # Inline text
|
|
157
|
-
|
|
158
|
-
# Isolation
|
|
159
|
-
zeroshot run 123 --worktree # git worktree
|
|
160
|
-
zeroshot run 123 --docker # container
|
|
161
|
-
|
|
162
|
-
# Automation (--ship implies --pr implies --worktree)
|
|
163
|
-
zeroshot run 123 --pr # worktree + create PR
|
|
164
|
-
zeroshot run 123 --ship # PR + auto-merge on approval
|
|
165
|
-
|
|
166
|
-
# Background mode
|
|
167
|
-
zeroshot run 123 -d
|
|
168
|
-
zeroshot run 123 --ship -d
|
|
169
|
-
|
|
170
|
-
# Control
|
|
171
|
-
zeroshot list
|
|
172
|
-
zeroshot status <id>
|
|
173
|
-
zeroshot logs <id> -f
|
|
174
|
-
zeroshot resume <id>
|
|
175
|
-
zeroshot stop <id>
|
|
176
|
-
zeroshot kill <id>
|
|
177
|
-
zeroshot watch
|
|
178
|
-
|
|
179
|
-
# Providers
|
|
180
|
-
zeroshot providers
|
|
181
|
-
zeroshot providers set-default codex
|
|
182
|
-
|
|
183
|
-
# Agent library
|
|
184
|
-
zeroshot agents list
|
|
185
|
-
zeroshot agents show <name>
|
|
186
|
-
|
|
187
|
-
# Maintenance
|
|
188
|
-
zeroshot clean
|
|
189
|
-
zeroshot purge
|
|
190
|
-
```
|
|
191
|
-
|
|
192
|
-
## Multi-Platform Issue Support
|
|
193
|
-
|
|
194
|
-
Zeroshot works with **GitHub, GitLab, Jira, and Azure DevOps**. Just paste the issue URL or key.
|
|
195
|
-
When working in a git repository, zeroshot automatically detects the issue provider from your git remote URL. No configuration needed!
|
|
196
|
-
|
|
197
|
-
```bash
|
|
198
|
-
# GitHub
|
|
199
|
-
zeroshot run 123
|
|
200
|
-
zeroshot run https://github.com/org/repo/issues/123
|
|
201
|
-
|
|
202
|
-
# GitLab (cloud and self-hosted)
|
|
203
|
-
zeroshot run https://gitlab.com/org/repo/-/issues/456
|
|
204
|
-
zeroshot run https://gitlab.mycompany.com/org/repo/-/issues/789
|
|
205
|
-
|
|
206
|
-
# Jira
|
|
207
|
-
zeroshot run PROJ-789
|
|
208
|
-
zeroshot run https://company.atlassian.net/browse/PROJ-789
|
|
209
|
-
|
|
210
|
-
# Azure DevOps
|
|
211
|
-
zeroshot run https://dev.azure.com/org/project/_workitems/edit/999
|
|
212
|
-
```
|
|
213
|
-
|
|
214
|
-
**Requires**: CLI tools ([`gh`](https://cli.github.com/), [`glab`](https://gitlab.com/gitlab-org/cli), [`jira`](https://github.com/go-jira/jira), or [`az`](https://docs.microsoft.com/cli/azure/)) for the platform you use. See [issue-providers README](src/issue-providers/README.md) for setup and self-hosted instances.
|
|
215
|
-
|
|
216
|
-
**Important for `--pr` mode**: Run zeroshot from the target repository directory. PRs are created on the git remote of your current directory. If you run from a different repo, zeroshot will warn you and skip the "Closes #X" reference (the PR is still created, but won't auto-close the issue).
|
|
217
|
-
|
|
218
|
-
## Architecture
|
|
219
|
-
|
|
220
|
-
Zeroshot is a message-driven coordination layer with smart defaults.
|
|
221
|
-
|
|
222
|
-
- The conductor classifies tasks by complexity and type.
|
|
223
|
-
- A workflow template selects agents and validators.
|
|
224
|
-
- Agents publish results to a SQLite ledger.
|
|
225
|
-
- Validators approve or reject with specific findings.
|
|
226
|
-
- Rejections route back to the worker for fixes.
|
|
227
|
-
|
|
228
|
-
```
|
|
229
|
-
┌─────────────────┐
|
|
230
|
-
│ TASK │
|
|
231
|
-
└────────┬────────┘
|
|
232
|
-
│
|
|
233
|
-
▼
|
|
234
|
-
┌────────────────────────────────────────────┐
|
|
235
|
-
│ CONDUCTOR │
|
|
236
|
-
│ Complexity × TaskType → Workflow │
|
|
237
|
-
└────────────────────────┬───────────────────┘
|
|
238
|
-
│
|
|
239
|
-
┌─────────────────────────────┼─────────────────────────────┐
|
|
240
|
-
│ │ │
|
|
241
|
-
▼ ▼ ▼
|
|
242
|
-
┌───────────┐ ┌───────────┐ ┌───────────┐
|
|
243
|
-
│ TRIVIAL │ │ SIMPLE │ │ STANDARD+ │
|
|
244
|
-
│ 1 agent │──────────▶ │ worker │ │ planner │
|
|
245
|
-
│ (level1) │ COMPLETE │ + 1 valid.│ │ + worker │
|
|
246
|
-
│ no valid. │ └─────┬─────┘ │ + 3-5 val.│
|
|
247
|
-
└───────────┘ │ └─────┬─────┘
|
|
248
|
-
▼ │
|
|
249
|
-
┌─────────────┐ ▼
|
|
250
|
-
┌──▶│ WORKER │ ┌─────────────┐
|
|
251
|
-
│ └──────┬──────┘ │ PLANNER │
|
|
252
|
-
│ │ └──────┬──────┘
|
|
253
|
-
│ ▼ │
|
|
254
|
-
│ ┌─────────────────────┐ ▼
|
|
255
|
-
│ │ ✓ validator │ ┌─────────────┐
|
|
256
|
-
│ │ (generic check) │ ┌──▶│ WORKER │
|
|
257
|
-
│ └──────────┬──────────┘ │ └──────┬──────┘
|
|
258
|
-
│ REJECT │ ALL OK │ │
|
|
259
|
-
└──────────────┘ │ │ ▼
|
|
260
|
-
│ │ ┌──────────────────────┐
|
|
261
|
-
│ │ │ ✓ requirements │
|
|
262
|
-
│ │ │ ✓ code (STANDARD+) │
|
|
263
|
-
│ │ │ ✓ security (CRIT) │
|
|
264
|
-
│ │ │ ✓ tester (CRIT) │
|
|
265
|
-
│ │ │ ✓ adversarial │
|
|
266
|
-
│ │ │ (real execution) │
|
|
267
|
-
│ │ └──────────┬───────────┘
|
|
268
|
-
│ │ REJECT │ ALL OK
|
|
269
|
-
│ └──────────────┘ │
|
|
270
|
-
▼ ▼
|
|
271
|
-
┌─────────────────────────────────────────────────────────────────────────────┐
|
|
272
|
-
│ COMPLETE │
|
|
273
|
-
└─────────────────────────────────────────────────────────────────────────────┘
|
|
274
|
-
```
|
|
275
|
-
|
|
276
|
-
### Complexity Model
|
|
277
|
-
|
|
278
|
-
| Task | Complexity | Agents | Validators |
|
|
279
|
-
| ---------------------- | ---------- | ------ | ------------------------------------------------- |
|
|
280
|
-
| Fix typo in README | TRIVIAL | 1 | None |
|
|
281
|
-
| Add dark mode toggle | SIMPLE | 2 | Generic validator |
|
|
282
|
-
| Refactor auth system | STANDARD | 4 | Requirements, code |
|
|
283
|
-
| Implement payment flow | CRITICAL | 7 | Requirements, code, security, tester, adversarial |
|
|
284
|
-
|
|
285
|
-
### Model Selection by Complexity
|
|
286
|
-
|
|
287
|
-
| Complexity | Planner | Worker | Validators |
|
|
288
|
-
| ---------- | ------- | ------ | ---------- |
|
|
289
|
-
| TRIVIAL | - | level1 | - |
|
|
290
|
-
| SIMPLE | - | level2 | 1 (level2) |
|
|
291
|
-
| STANDARD | level2 | level2 | 2 (level2) |
|
|
292
|
-
| CRITICAL | level3 | level2 | 5 (level2) |
|
|
293
|
-
|
|
294
|
-
Levels map to provider-specific models. Configure with `zeroshot providers setup <provider>` or
|
|
295
|
-
`settings.providerSettings`. (Legacy `maxModel` applies to Claude only.)
|
|
296
|
-
|
|
297
|
-
<details>
|
|
298
|
-
<summary><strong>Custom Workflows (Framework Mode)</strong></summary>
|
|
299
|
-
|
|
300
|
-
Zeroshot is message-driven, so you can define any agent topology.
|
|
301
|
-
|
|
302
|
-
- Expert panels: parallel specialists -> aggregator -> decision
|
|
303
|
-
- Staged gates: sequential validators, each with veto power
|
|
304
|
-
- Hierarchical: supervisor dynamically spawns workers
|
|
305
|
-
- Dynamic: conductor adds agents mid-execution
|
|
306
|
-
|
|
307
|
-
**Coordination primitives:**
|
|
308
|
-
|
|
309
|
-
- Message bus (pub/sub topics)
|
|
310
|
-
- Triggers (wake agents on conditions)
|
|
311
|
-
- Ledger (SQLite, crash recovery)
|
|
312
|
-
- Dynamic spawning (CLUSTER_OPERATIONS)
|
|
313
|
-
|
|
314
|
-
#### Creating Custom Clusters with a Provider CLI
|
|
315
|
-
|
|
316
|
-
Start your provider CLI and describe your cluster:
|
|
317
|
-
|
|
318
|
-
```
|
|
319
|
-
Create a zeroshot cluster config for security-critical features:
|
|
320
|
-
|
|
321
|
-
1. Implementation agent (level2) implements the feature
|
|
322
|
-
2. FOUR parallel validators:
|
|
323
|
-
- Security validator: OWASP checks, SQL injection, XSS, CSRF
|
|
324
|
-
- Performance validator: No N+1 queries, proper indexing
|
|
325
|
-
- Privacy validator: GDPR compliance, data minimization
|
|
326
|
-
- Code reviewer: General code quality
|
|
327
|
-
|
|
328
|
-
3. ALL validators must approve before merge
|
|
329
|
-
4. If ANY validator rejects, implementation agent fixes and resubmits
|
|
330
|
-
5. Use level3 for security validator (highest stakes)
|
|
331
|
-
|
|
332
|
-
Look at cluster-templates/base-templates/full-workflow.json
|
|
333
|
-
and create a similar cluster. Save to cluster-templates/security-review.json
|
|
334
|
-
```
|
|
335
|
-
|
|
336
|
-
Built-in validation checks for missing triggers, deadlocks, and invalid type wiring before running.
|
|
337
|
-
|
|
338
|
-
See [CLAUDE.md](./CLAUDE.md) for the cluster schema and examples.
|
|
339
|
-
|
|
340
|
-
</details>
|
|
341
|
-
|
|
342
|
-
## Crash Recovery
|
|
343
|
-
|
|
344
|
-
All state is persisted in the SQLite ledger. You can resume at any time:
|
|
345
|
-
|
|
346
|
-
```bash
|
|
347
|
-
zeroshot resume cluster-bold-panther
|
|
348
|
-
```
|
|
349
|
-
|
|
350
|
-
## Isolation Modes
|
|
351
|
-
|
|
352
|
-
### Git Worktree (Default for --pr/--ship)
|
|
5
|
+
Install the maintained package directly:
|
|
353
6
|
|
|
354
7
|
```bash
|
|
355
|
-
|
|
8
|
+
npm install -g @the-open-engine/zeroshot
|
|
356
9
|
```
|
|
357
10
|
|
|
358
|
-
|
|
359
|
-
|
|
360
|
-
### Docker Container
|
|
11
|
+
If this bridge package is already installed, run:
|
|
361
12
|
|
|
362
13
|
```bash
|
|
363
|
-
zeroshot
|
|
14
|
+
zeroshot update
|
|
364
15
|
```
|
|
365
16
|
|
|
366
|
-
|
|
367
|
-
|
|
368
|
-
### When to Use Which
|
|
369
|
-
|
|
370
|
-
| Scenario | Recommended |
|
|
371
|
-
| ------------------------------------ | ---------------------- |
|
|
372
|
-
| Quick task, review changes yourself | No isolation (default) |
|
|
373
|
-
| PR workflow, code review | `--worktree` or `--pr` |
|
|
374
|
-
| Risky experiment, might break things | `--docker` |
|
|
375
|
-
| Running multiple tasks in parallel | `--docker` |
|
|
376
|
-
| Full automation, no review needed | `--ship` |
|
|
377
|
-
|
|
378
|
-
**Default behavior:** Agents modify files only; they do not commit or push unless using an isolation mode that explicitly allows it.
|
|
379
|
-
|
|
380
|
-
<details>
|
|
381
|
-
<summary><strong>Docker Credential Mounts</strong></summary>
|
|
382
|
-
|
|
383
|
-
When using `--docker`, zeroshot mounts credential directories so agents can access provider CLIs and tools like AWS, Azure, and kubectl.
|
|
384
|
-
|
|
385
|
-
**Default mounts**: `gh`, `git`, `ssh` (GitHub CLI, git config, SSH keys)
|
|
386
|
-
|
|
387
|
-
**Available presets**: `gh`, `git`, `ssh`, `aws`, `azure`, `kube`, `terraform`, `gcloud`, `claude`, `codex`, `gemini`
|
|
388
|
-
|
|
389
|
-
```bash
|
|
390
|
-
# Configure via settings (persistent)
|
|
391
|
-
zeroshot settings set dockerMounts '["gh", "git", "ssh", "aws", "azure"]'
|
|
392
|
-
|
|
393
|
-
# View current config
|
|
394
|
-
zeroshot settings get dockerMounts
|
|
395
|
-
|
|
396
|
-
# Per-run override
|
|
397
|
-
zeroshot run 123 --docker --mount ~/.aws:/root/.aws:ro
|
|
398
|
-
|
|
399
|
-
# Provider credentials
|
|
400
|
-
zeroshot run 123 --docker --mount ~/.config/codex:/home/node/.config/codex:ro
|
|
401
|
-
zeroshot run 123 --docker --mount ~/.config/gemini:/home/node/.config/gemini:ro
|
|
402
|
-
|
|
403
|
-
# Disable all mounts
|
|
404
|
-
zeroshot run 123 --docker --no-mounts
|
|
405
|
-
|
|
406
|
-
# CI: env var override
|
|
407
|
-
ZEROSHOT_DOCKER_MOUNTS='["aws","azure"]' zeroshot run 123 --docker
|
|
408
|
-
```
|
|
409
|
-
|
|
410
|
-
See `docs/providers.md` for provider CLI setup and mount details.
|
|
411
|
-
|
|
412
|
-
**Custom mounts** (mix presets with explicit paths):
|
|
413
|
-
|
|
414
|
-
```bash
|
|
415
|
-
zeroshot settings set dockerMounts '[
|
|
416
|
-
"gh",
|
|
417
|
-
"git",
|
|
418
|
-
{"host": "~/.myconfig", "container": "$HOME/.myconfig", "readonly": true}
|
|
419
|
-
]'
|
|
420
|
-
```
|
|
421
|
-
|
|
422
|
-
**Container home**: Presets use `$HOME` placeholder. Default: `/root`. Override with:
|
|
423
|
-
|
|
424
|
-
```bash
|
|
425
|
-
zeroshot settings set dockerContainerHome '/home/node'
|
|
426
|
-
# Or per-run:
|
|
427
|
-
zeroshot run 123 --docker --container-home /home/node
|
|
428
|
-
```
|
|
429
|
-
|
|
430
|
-
**Env var passthrough**: Presets auto-pass related env vars (for example, `aws` -> `AWS_REGION`, `AWS_PROFILE`). Add custom:
|
|
431
|
-
|
|
432
|
-
```bash
|
|
433
|
-
zeroshot settings set dockerEnvPassthrough '["MY_API_KEY", "TF_VAR_*"]'
|
|
434
|
-
```
|
|
435
|
-
|
|
436
|
-
</details>
|
|
437
|
-
|
|
438
|
-
## Resources
|
|
439
|
-
|
|
440
|
-
- [CLAUDE.md](./CLAUDE.md) - Architecture, cluster config schema, agent primitives
|
|
441
|
-
- `docs/providers.md` - Provider setup, model levels, and Docker mounts
|
|
442
|
-
- [Discord](https://discord.gg/PdZ3UEXB) - Support and community
|
|
443
|
-
- `zeroshot export <id>` - Export conversation to markdown
|
|
444
|
-
- `sqlite3 ~/.zeroshot/*.db` - Direct ledger access for debugging
|
|
445
|
-
|
|
446
|
-
<details>
|
|
447
|
-
<summary><strong>Troubleshooting</strong></summary>
|
|
448
|
-
|
|
449
|
-
| Issue | Fix |
|
|
450
|
-
| ----------------------------- | ----------------------------------------------------------------------------------------- |
|
|
451
|
-
| `claude: command not found` | `npm i -g @anthropic-ai/claude-code && claude auth login` |
|
|
452
|
-
| `codex: command not found` | `npm i -g @openai/codex && codex login` |
|
|
453
|
-
| `gemini: command not found` | `npm i -g @google/gemini-cli && gemini auth login` |
|
|
454
|
-
| `gh: command not found` | [Install GitHub CLI](https://cli.github.com/) |
|
|
455
|
-
| `--docker` fails | Docker must be running: `docker ps` to verify |
|
|
456
|
-
| Cluster stuck | `zeroshot resume <id>` to continue |
|
|
457
|
-
| Agent keeps failing | Check `zeroshot logs <id>` for actual error |
|
|
458
|
-
| `zeroshot: command not found` | `npm install -g @covibes/zeroshot` |
|
|
459
|
-
| Agents misbehave | `/analyze-cluster-postmortem <id>` in Claude Code (creates issue if fix is generalizable) |
|
|
460
|
-
|
|
461
|
-
</details>
|
|
462
|
-
|
|
463
|
-
## Contributing
|
|
464
|
-
|
|
465
|
-
See [CONTRIBUTING.md](CONTRIBUTING.md) for development setup and guidelines.
|
|
466
|
-
|
|
467
|
-
Please read [CODE_OF_CONDUCT.md](CODE_OF_CONDUCT.md) before participating.
|
|
468
|
-
|
|
469
|
-
For security issues, see [SECURITY.md](SECURITY.md).
|
|
470
|
-
|
|
471
|
-
---
|
|
472
|
-
|
|
473
|
-
MIT - [Covibes](https://github.com/covibes)
|
|
474
|
-
|
|
475
|
-
Built on [Claude Code](https://claude.com/product/claude-code) by Anthropic.
|
|
17
|
+
The bridge prints a migration notice on every CLI invocation and delegates normal commands to
|
|
18
|
+
`@the-open-engine/zeroshot`.
|