infer
Use the infer skill to generate speculative inference candidates from claims, persist them on branch-local inference graphs, and run the full mutation pipeline that coordinates active memory, claims, Terminus persistence, and speculative ranking.
来源信息
- 仓库
- Tyler-R-Kendrick/agentic_speculative_knowledge
- 最近来源活动
- 2026年4月10日 14:25
- 检测到的 SKILL.md 语言
- 英语
- 星标
- 0
- 分支
- 0
安装方式
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
检查来源文件
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
正在显示 SKILL.md
SKILL.md
来源说明 · 只读预览- name
- infer
- description
- Use the infer skill to generate speculative inference candidates from claims, persist them on branch-local inference graphs, and run the full mutation pipeline that coordinates active memory, claims, Terminus persistence, and speculative ranking.
# Infer
Generate speculative knowledge from existing claims and persist it on
isolated inference branches.
## When to use
- Running the mutation pipeline end-to-end on new observations
- Generating inference candidates (abductive, heuristic, predictive, etc.)
- Producing facet relation candidates between claims
- Persisting ranked speculative output to a Terminus inference branch
## Quick start
```python
import pathlib
from src.active_memory.models import WorkingItem
from src.persistence.pipeline import MutationPipeline
from src.terminus.adapter import TerminusMemoryRepository
from src.manifold_sidecar import ManifoldRankingService
root = pathlib.Path(".agent-memory")
repo = TerminusMemoryRepository(url="http://localhost:6363")
ranker = ManifoldRankingService()
pipeline = MutationPipeline(
root,
enable_terminus=True,
terminus_repo=repo,
manifold_service=ranker,
enable_inference=True,
)
item = WorkingItem(
item_type="observation",
content="The auth service failed after a certificate rotation.",
session_id="sess-1",
)
result = pipeline.run(item, session_id="sess-1")
print(result.inference_candidates) # number of candidates generated
print(result.ranked_inference_candidates) # number ranked by manifold sidecar
print(result.inference_branch) # e.g. "inference/sess-1"
```
### Using the generator directly
```python
from src.inference.generator import InferenceGenerator
from src.claims.extractor import ClaimExtractor
extractor = ClaimExtractor()
claims = extractor.extract(text="The API started returning 401 errors.")
generator = InferenceGenerator()
candidates = generator.generate_from_claims(
claims,
source_branch="inference/sess-1",
source_commit="journal-event-abc",
session_id="sess-1",
)
facets = generator.generate_facet_candidates(
claims,
provenance_commit="journal-event-abc",
source_branch="inference/sess-1",
)
```
## Key APIs
| Class | Method | Purpose |
|---|---|---|
| `MutationPipeline` | `run()` | Full pipeline: active write → journal → claims → Terminus → inference → ranking |
| `InferenceGenerator` | `generate_from_claims()` | Produce `InferenceNode` candidates from a list of claims |
| `InferenceGenerator` | `generate_facet_candidates()` | Produce `FacetRelation` candidates between adjacent claims |
## Grounding
- `src/persistence/pipeline.py` — mutation pipeline orchestration
- `src/inference/generator.py` — rule-based inference and facet generation
- `src/inference/models.py` — `InferenceNode` and `FacetRelation` schemas
- `src/claims/extractor.py` — upstream claim extraction
- `tests/integration/test_inference_flow.py` — integration coverage
- `notebooks/02_speculative_inference_and_facets.ipynb` — executable walkthrough
## Rules
1. Route orchestration through `MutationPipeline` in `src/persistence/pipeline.py` so all pipeline steps stay synchronized.
2. Reuse `InferenceGenerator` from `src/inference/generator.py` for candidate creation.
3. Speculative output must stay on `inference/*` branches; never write directly to trusted memory.
4. Validate with `tests/integration/test_inference_flow.py`.
在 GitHub 查看