scriptoriumadmin/ReMe
1
0
Fork 0
mirror of https://github.com/agentscope-ai/ReMe.git synced 2026-10-23 05:35:30 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 9

docs: describe current decision model configuration directly

Browse source
This commit is contained in:
jinli.yl 2026-10-15 16:38:54 +08:00
parent 4d43236db2
commit 62056c14ad
4 changed files with 4 additions and 16 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

6
docs/en/configuration.md
View file

@ -198,12 +198,6 @@ Implement `BaseAsDecisionModel` and register a backend to add another provider.
they do not select a provider directly. The current adapter uses AgentScope classifier question and answer schemas.
See `reme/components/as_decision_model/` for the interface. A null `model` uses the adapter default.
Migration: rename `as_classifier` to `as_decision_model`, backend `jev` to `agentscope`, config `classifier_demo` to
`decision_demo`, Job `jev` to `decide`, and install extra `classifier-jev` to `decision`. Environment variables now use
`DECISION_MODEL_BACKEND`, `DECISION_MODEL_NAME`, `DECISION_MODEL_API_KEY`, and `DECISION_MODEL_BASE_URL`. Python imports
move to `reme.components.as_decision_model`; the base and adapter are `BaseAsDecisionModel` and
`AgentScopeAsDecisionModel`. Step output is available as `context.data["decision_result"]`. Old names are not aliases.
For tool-history compaction, start `reme start config=decision_compaction` and call `reme compact_context` from another
terminal with `messages` and any protection or budget options. It uses the same decision model interface and returns
a proposed snapshot; the caller owns archiving and applying it. Request token budgets are estimates with overhead

4
docs/en/context_compaction.md
View file

@ -65,8 +65,8 @@ export DECISION_MODEL_BACKEND=agentscope
# Optional: DECISION_MODEL_NAME and DECISION_MODEL_BASE_URL
```
The CLI also loads `.env`. Migrate old `CLASSIFIER_*` variables to `DECISION_MODEL_*` as described in
[configuration](./configuration.md#decision-model-configuration); the backend name is `agentscope`.
The CLI also loads these `DECISION_MODEL_*` variables from `.env`. The backend name is `agentscope`; see
[decision model configuration](./configuration.md#decision-model-configuration) for the complete settings.
Omitting the model name or setting it to `null` uses the adapter's SDK default.
Start the service in one terminal:

6
docs/zh/configuration.md
View file

@ -196,12 +196,6 @@ SDK 错误沿用 BaseJob 的 `success=false` 契约。
新增供应商时，实现 `BaseAsDecisionModel` 并注册 backend；Job 按组件名选择模型。当前适配器使用 AgentScope classifier
schema 定义问题与答案。接口见 `reme/components/as_decision_model/`。`model` 为 null 时使用适配器默认模型。
迁移：组件键 `as_classifier` 改为 `as_decision_model`，backend `jev` 改为 `agentscope`，配置 `classifier_demo` 改为
`decision_demo`，Job `jev` 改为 `decide`，安装 extra `classifier-jev` 改为 `decision`。环境变量改为
`DECISION_MODEL_BACKEND`、`DECISION_MODEL_NAME`、`DECISION_MODEL_API_KEY` 和 `DECISION_MODEL_BASE_URL`。Python 导入路径改为
`reme.components.as_decision_model`，基类和适配器分别为 `BaseAsDecisionModel`、`AgentScopeAsDecisionModel`；Step 输出键改为
`context.data["decision_result"]`。旧名称不保留别名。
工具历史压缩使用同一决策模型接口：先启动 `reme start config=decision_compaction`，再在另一终端调用
`reme compact_context`，传入 `messages` 以及需要的保护和预算选项。返回值是建议的快照，归档和应用由调用方负责。
请求 token 预算使用预留开销的估算值，与宿主模型的 tokenizer 独立。

4
docs/zh/context_compaction.md
View file

@ -55,8 +55,8 @@ export DECISION_MODEL_BACKEND=agentscope
# 可选：DECISION_MODEL_NAME、DECISION_MODEL_BASE_URL
```
`.env` 也可以配置这些变量。旧的 `CLASSIFIER_*` 名称需按[配置迁移说明](./configuration.md#决策模型配置)改为
`DECISION_MODEL_*`；backend 应为 `agentscope`。省略模型名或配置为 `null` 时使用适配器的 SDK 默认模型。
`.env` 也可以配置上述 `DECISION_MODEL_*` 变量，backend 使用 `agentscope`，完整设置见
[决策模型配置](./configuration.md#决策模型配置)。省略模型名或配置为 `null` 时使用适配器的 SDK 默认模型。
在一个终端启动服务：