debaitable 0.0.1
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/README.md +62 -0
- package/dist/main.js +2212 -0
- package/package.json +34 -0
package/README.md
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
1
|
+
# DebAItable
|
|
2
|
+
|
|
3
|
+
DebAItable is an artifact-first multi-agent decision engine. Instead of free-form chat, it runs
|
|
4
|
+
structured debate rounds and outputs a Decision Record plus an audit trail of all role outputs.
|
|
5
|
+
|
|
6
|
+
## What It Does
|
|
7
|
+
- Runs a 3-round debate pipeline:
|
|
8
|
+
- Round 1: independent proposals per role
|
|
9
|
+
- Round 2: critiques and rebuttals
|
|
10
|
+
- Round 3: convergence vote
|
|
11
|
+
- Produces a structured `DecisionRecord`:
|
|
12
|
+
- summary, rationale, tradeoffs, risks, actions, confidence, minority report
|
|
13
|
+
- executive decision block (`decision`, `why`, top risks/actions, stop/go criteria)
|
|
14
|
+
- Stores all rounds and run states in a persistence layer
|
|
15
|
+
- Supports OpenAI-backed generation and deterministic local heuristic mode
|
|
16
|
+
- Includes an interactive TUI CLI that saves JSON artifacts under `artifacts/`
|
|
17
|
+
- Uses one-message decision intake by default, with optional guided edit only when needed
|
|
18
|
+
- Applies normalization + quality gates to keep output concise, non-duplicative, and usable
|
|
19
|
+
|
|
20
|
+
## Current Architecture
|
|
21
|
+
1. API service validates and sanitizes decision input.
|
|
22
|
+
2. Decision is persisted and enqueued as a background-style job payload.
|
|
23
|
+
3. Job runner executes orchestration rounds via role prompts.
|
|
24
|
+
4. Outputs are schema-validated with Zod at each boundary.
|
|
25
|
+
5. Final record and full round audit trail are stored and returned.
|
|
26
|
+
|
|
27
|
+
## Run It
|
|
28
|
+
1. Install dependencies:
|
|
29
|
+
- `npm install`
|
|
30
|
+
2. Optional environment:
|
|
31
|
+
- `OPENAI_API_KEY` (if omitted, CLI uses deterministic heuristic provider)
|
|
32
|
+
- `OPENAI_MODEL` (default: `gpt-5`)
|
|
33
|
+
3. Run demo:
|
|
34
|
+
- `npm run dev`
|
|
35
|
+
4. Run interactive TUI:
|
|
36
|
+
- `npm run cli`
|
|
37
|
+
- Keyboard-first full-screen workflow:
|
|
38
|
+
- Type a prompt/question in the left input pane and press `Enter` to run
|
|
39
|
+
- Arrow keys move focus between prompt/history/output panes
|
|
40
|
+
- `a` toggles audit timeline
|
|
41
|
+
- `m` switches model mode (OpenAI/heuristic when API key is present)
|
|
42
|
+
- `?` opens compact help
|
|
43
|
+
5. Typecheck:
|
|
44
|
+
- `npm run typecheck`
|
|
45
|
+
|
|
46
|
+
## Install From npm (Global CLI)
|
|
47
|
+
1. Publish package from this repo:
|
|
48
|
+
- `npm login`
|
|
49
|
+
- `npm publish --access public`
|
|
50
|
+
2. Install globally:
|
|
51
|
+
- `npm i -g debaitable`
|
|
52
|
+
3. Run from any terminal:
|
|
53
|
+
- `debaitable`
|
|
54
|
+
|
|
55
|
+
## Repository Scope (Today)
|
|
56
|
+
- Engine modules under `src/`:
|
|
57
|
+
- `core`, `api`, `orchestration`, `jobs`, `persistence`, `ai`, `cli`
|
|
58
|
+
- In-memory queue and store are the default runtime adapters.
|
|
59
|
+
- No web UI or external DB wiring in this repo yet.
|
|
60
|
+
|
|
61
|
+
## License
|
|
62
|
+
All rights reserved. No license is granted to use, copy, modify, or distribute this software.
|