edge-pipeline-orchestrator

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Edge Pipeline Orchestrator

Edge流水线编排器

Coordinate all edge research stages into a single automated pipeline run.

将所有边缘研究阶段协调为单个自动化流水线运行。

When to Use

适用场景

Run the full edge pipeline from tickets (or OHLCV) to exported strategies
Resume a partially completed pipeline from the drafts stage
Review and revise existing strategy drafts with feedback loop
Dry-run the pipeline to preview results without exporting

运行从工单（或OHLCV）到导出策略的全Edge流水线
从草稿阶段恢复部分完成的流水线
通过反馈回路审核和修订现有策略草稿
试运行流水线以预览结果，无需导出

Workflow

工作流

Load pipeline configuration from CLI arguments
Run auto_detect stage if --from-ohlcv is provided (generates tickets from raw OHLCV data)
Run hints stage to extract edge hints from market summary and anomalies
Run concepts stage to synthesize abstract edge concepts from tickets and hints
Run drafts stage to design strategy drafts from concepts
Run review-revision feedback loop:
- Review all drafts (max 2 iterations)
- PASS verdicts accumulated; REJECT verdicts accumulated
- REVISE verdicts trigger apply_revisions and re-review
- Remaining REVISE after max iterations downgraded to research_probe
Export eligible drafts (PASS + export_ready_v1 + exportable entry_family)
Write pipeline_run_manifest.json with full execution trace

从CLI参数加载流水线配置
如果提供了--from-ohlcv参数则运行auto_detect阶段（从原始OHLCV数据生成工单）
运行hints阶段，从市场汇总和异常数据中提取edge提示
运行concepts阶段，用工单和提示合成抽象edge概念
运行drafts阶段，基于概念设计策略草稿
运行审核-修订反馈回路：
- 审核所有草稿（最多2次迭代）
- 累计PASS结论；累计REJECT结论
- REVISE结论触发apply_revisions并重新审核
- 达到最大迭代次数后剩余的REVISE结论降级为research_probe
导出符合条件的草稿（PASS + export_ready_v1 + 可导出的entry_family）
写入包含完整执行轨迹的pipeline_run_manifest.json

CLI Usage

CLI用法

bash

undefined

bash

undefined

Full pipeline from tickets

python3 scripts/orchestrate_edge_pipeline.py
--tickets-dir path/to/tickets/
--output-dir reports/edge_pipeline/

Full pipeline from OHLCV

python3 scripts/orchestrate_edge_pipeline.py
--from-ohlcv path/to/ohlcv.csv
--output-dir reports/edge_pipeline/

Resume from drafts stage

python3 scripts/orchestrate_edge_pipeline.py
--resume-from drafts
--drafts-dir path/to/drafts/
--output-dir reports/edge_pipeline/

Review-only mode

python3 scripts/orchestrate_edge_pipeline.py
--review-only
--drafts-dir path/to/drafts/
--output-dir reports/edge_pipeline/

Dry run (no export)

python3 scripts/orchestrate_edge_pipeline.py
--tickets-dir path/to/tickets/
--output-dir reports/edge_pipeline/
--dry-run

undefined

python3 scripts/orchestrate_edge_pipeline.py
--tickets-dir path/to/tickets/
--output-dir reports/edge_pipeline/
--dry-run

undefined

Output

输出

All artifacts are written to

--output-dir

output-dir/
├── pipeline_run_manifest.json
├── tickets/          (from auto_detect)
├── hints/hints.yaml  (from hints)
├── concepts/edge_concepts.yaml
├── drafts/*.yaml
├── exportable_tickets/*.yaml
├── reviews_iter_0/*.yaml
├── reviews_iter_1/*.yaml  (if needed)
└── strategies/<candidate_id>/
    ├── strategy.yaml
    └── metadata.json

所有产物都会写入到

--output-dir

指定的目录：

output-dir/
├── pipeline_run_manifest.json
├── tickets/          (from auto_detect)
├── hints/hints.yaml  (from hints)
├── concepts/edge_concepts.yaml
├── drafts/*.yaml
├── exportable_tickets/*.yaml
├── reviews_iter_0/*.yaml
├── reviews_iter_1/*.yaml  (if needed)
└── strategies/<candidate_id>/
    ├── strategy.yaml
    └── metadata.json

Claude Code LLM-Augmented Workflow

Claude Code LLM增强工作流

Run the LLM-augmented pipeline entirely within Claude Code:

Run auto_detect to produce
```
market_summary.json
```
+
```
anomalies.json
```
Claude Code analyzes data and generates edge hints
Save hints to a YAML file:

yaml

- title: Sector rotation into industrials
  observation: Tech underperforming while industrials show relative strength
  symbols: [CAT, DE, GE]
  regime_bias: Neutral
  mechanism_tag: flow
  preferred_entry_family: pivot_breakout
  hypothesis_type: sector_x_stock

Run orchestrator with
```
--llm-ideas-file
```
and
```
--promote-hints
```
:

bash

python3 scripts/orchestrate_edge_pipeline.py \
  --tickets-dir path/to/tickets/ \
  --llm-ideas-file llm_hints.yaml \
  --promote-hints \
  --as-of 2026-02-28 \
  --max-synthetic-ratio 1.5 \
  --strict-export \
  --output-dir reports/edge_pipeline/

完全在Claude Code中运行LLM增强的流水线：

运行auto_detect生成
```
market_summary.json
```
+
```
anomalies.json
```
Claude Code分析数据并生成edge提示
将提示保存到YAML文件：

yaml

- title: Sector rotation into industrials
  observation: Tech underperforming while industrials show relative strength
  symbols: [CAT, DE, GE]
  regime_bias: Neutral
  mechanism_tag: flow
  preferred_entry_family: pivot_breakout
  hypothesis_type: sector_x_stock

使用
```
--llm-ideas-file
```
和
```
--promote-hints
```
参数运行编排器：

bash

python3 scripts/orchestrate_edge_pipeline.py \
  --tickets-dir path/to/tickets/ \
  --llm-ideas-file llm_hints.yaml \
  --promote-hints \
  --as-of 2026-02-28 \
  --max-synthetic-ratio 1.5 \
  --strict-export \
  --output-dir reports/edge_pipeline/

Optional Flags

可选参数

```
--as-of YYYY-MM-DD
```
— forwarded to hints stage for date filtering
```
--strict-export
```
— export-eligible drafts with any warn finding get REVISE instead of PASS
```
--max-synthetic-ratio N
```
— cap synthetic tickets to N × real ticket count (floor: 3)
```
--overlap-threshold F
```
— condition overlap threshold for concept deduplication (default: 0.75)
```
--no-dedup
```
— disable concept deduplication

Note:

--llm-ideas-file

and

--promote-hints

are effective only during full pipeline runs.

--resume-from drafts

and

--review-only

skip hints/concepts stages, so these flags are ignored.

```
--as-of YYYY-MM-DD
```
— 传递到hints阶段用于日期过滤
```
--strict-export
```
— 存在任何警告的符合导出条件的草稿将被标记为REVISE而非PASS
```
--max-synthetic-ratio N
```
— 限制合成工单数量为真实工单数量的N倍（下限：3）
```
--overlap-threshold F
```
— 概念去重的条件重叠阈值（默认：0.75）
```
--no-dedup
```
— 禁用概念去重

注意：

--llm-ideas-file

和

--promote-hints

仅在全流水线运行时生效。

--resume-from drafts

和

--review-only

会跳过hints/concepts阶段，因此这些参数会被忽略。

Resources

资源

```
references/pipeline_flow.md
```
— Pipeline stages, data contracts, and architecture
```
references/revision_loop_rules.md
```
— Review-revision feedback loop rules and heuristics

```
references/pipeline_flow.md
```
— 流水线阶段、数据契约和架构说明
```
references/revision_loop_rules.md
```
— 审核-修订反馈回路规则和启发式方法