1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
//! Attractor Mode — DOT-based AI workflow graph execution.
//!
//! Attractor Mode lets you define multi-step AI pipelines as DOT graphs
//! where nodes represent tasks (LLM calls, human gates, tool invocations)
//! and edges represent transitions with optional conditions.
//!
//! # Architecture
//!
//! ```text
//! DOT file → Parser → PipelineGraph → Transforms → Validator → Runner
//! ↓
//! AgentBackend ← Handler Registry
//! ```
//!
//! # Modules
//!
//! - [`dot_parser`] — Parse DOT subset into raw graph structures
//! - [`graph`] — Internal petgraph-based pipeline representation
//! - [`context`] — Thread-safe key-value execution context
//! - [`checkpoint`] — Checkpoint save/load for resumption
//! - [`run_directory`] — Filesystem layout for pipeline runs
//! - [`outcome`] — Stage execution outcomes and status
//! - [`runner`] — Core execution engine with edge selection
//! - [`conditions`] — Condition expression parser/evaluator
//! - [`retry`] — Retry policies with exponential backoff
//! - [`events`] — Pipeline execution events for observability
//! - [`handlers`] — Node handler trait and implementations
//! - [`interviewer`] — Human-in-the-loop interaction trait
//! - [`transforms`] — Pre-execution graph transforms
//! - [`stylesheet`] — CSS-like model/provider styling
//! - [`validator`] — Graph lint rules and validation