reflect
Use the reflect skill to persist memories, claims, and inference nodes into the TerminusDB temporal graph. Covers branch lifecycle, trusted session writes, temporal recall, and the separation between session and inference branches.
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
- reflect
- description
- Use the reflect skill to persist memories, claims, and inference nodes into the TerminusDB temporal graph. Covers branch lifecycle, trusted session writes, temporal recall, and the separation between session and inference branches.
# Reflect
Persist and reason over long-term knowledge in the temporal graph.
## When to use
- Writing memories or claims to a persistent Terminus branch
- Creating or managing session and inference branches
- Querying the temporal graph for trusted or branch-local data
- Verifying the separation between trusted and speculative branches
## Quick start
```python
from src.terminus.adapter import TerminusMemoryRepository
from src.terminus.branch_manager import (
BranchManager,
session_branch_name,
inference_branch_name,
)
repo = TerminusMemoryRepository(url="http://localhost:6363")
# Create a session branch
branch = repo.ensure_branch(session_branch_name("sess-1"))
# Write a claim to the session branch
from src.claims.models import Claim
claim = Claim(claim_text="Cert rotation caused auth failures.", claim_type="observation")
repo.write_claim(branch, claim)
# Write a memory
from src.normalization.mapper import ClaimToMemoryMapper
mapper = ClaimToMemoryMapper()
memories = mapper.map_many([claim], session_id="sess-1")
for m in memories:
m.source_branch = branch
repo.write_memory(branch, m)
# Query trusted memories
trusted = repo.query_memories(branch="session/sess-1")
# Query inference nodes on a speculative branch
inference_br = repo.ensure_branch(inference_branch_name("sess-1"))
nodes = repo.query_inference_nodes(inference_br)
```
### Branch manager helpers
```python
from src.terminus.branch_manager import session_branch_name, inference_branch_name
session_branch_name("sess-1") # → "session/sess-1"
inference_branch_name("sess-1") # → "inference/sess-1"
```
## Key APIs
| Class | Method | Purpose |
|---|---|---|
| `TerminusMemoryRepository` | `ensure_branch()` | Create a branch if it does not exist |
| `TerminusMemoryRepository` | `write_memory()` | Persist a `CandidateMemory` to a branch |
| `TerminusMemoryRepository` | `write_claim()` | Persist a `Claim` to a branch |
| `TerminusMemoryRepository` | `write_inference_node()` | Persist an `InferenceNode` to a branch |
| `TerminusMemoryRepository` | `write_facet_relation()` | Persist a `FacetRelation` to a branch |
| `TerminusMemoryRepository` | `query_memories()` | Read memories from a branch |
| `TerminusMemoryRepository` | `query_inference_nodes()` | Read inference nodes from a branch |
| `TerminusMemoryRepository` | `query_facet_relations()` | Read facet relations from a branch |
| `session_branch_name()` | — | `"session/{id}"` naming helper |
| `inference_branch_name()` | — | `"inference/{id}"` naming helper |
| `BranchManager` | `create_session_branch()` / `create_inference_branch()` | High-level branch creation |
## Grounding
- `src/terminus/adapter.py` — Terminus persistence adapter with fallback store
- `src/terminus/branch_manager.py` — branch naming and lifecycle helpers
- `src/terminus/schema.py` — document encoding/decoding for Terminus
- `tests/unit/test_terminus.py` — unit coverage
- `notebooks/04_terminus_temporal_graph_memory.ipynb` — executable walkthrough
## Rules
1. Use `TerminusMemoryRepository` from `src/terminus/adapter.py` for all graph writes and reads.
2. Reuse `session_branch_name()` and `inference_branch_name()` from `src/terminus/branch_manager.py` for branch naming.
3. Keep trusted data on `session/*` branches and speculative data on `inference/*` branches; never mix them.
4. The adapter falls back to an in-process store when Terminus is unreachable — keep both paths aligned.
5. Validate with `tests/unit/test_terminus.py`.
在 GitHub 查看