feat: system prompt playbook variable interpolation - #578
Open
angri450 wants to merge 2 commits into
Open
Conversation
Render {agent_id}/{agent_name}/{date}/{datetime}/{work_dir}/{model} in an
agent's system prompt when the harness graph is compiled, so a single
persona template stays accurate across instances (inspired by
openresearch-cli's playbook substitution). Unknown placeholders are left
untouched so template typos stay visible instead of silently vanishing.
- _render_system_prompt: module-level renderer in AgentManager
- _build_harness_config: render non-empty system_prompt after bootstrap;
resolve default_model once and reuse for the render + config
- test: render path incl. unknown-placeholder preservation
- CHANGELOG: [Unreleased] entry
There was a problem hiding this comment.
🟡 Changes recommended
system_prompt 的 {date}/{datetime} 渲染目前使用本机时区而非配置的 default_timezone，可能导致线上时间不一致且相关测试断言也偏弱。
Once you've addressed the issues Copilot identified, you can request another Copilot review.
Pull request overview
该 PR 为 Octop 的智能体 system_prompt（agents.persona）引入“playbook 风格”的模板变量插值，在编译 harness 图（graph-compile）阶段将 agent 身份与环境信息渲染进最终 system prompt，以便同一 persona 模板在不同 agent 实例间复用且保持上下文准确。
Changes:
- 在
AgentManager._build_harness_config中对非空system_prompt做{agent_id}/{agent_name}/{date}/{datetime}/{work_dir}/{model}渲染，并将default_model解析结果复用。 - 新增单测覆盖：渲染路径 + 未知占位符保留（
{foo}不被吞掉）。 - 更新 CHANGELOG 记录该能力（Unreleased）。
File summaries
| File | Description |
|---|---|
src/octop/infra/agents/manager.py |
新增 system_prompt 插值渲染函数，并在 build harness config 时应用渲染与复用 default_model |
tests/unit/agents/test_agent_manager.py |
增加 system_prompt 插值渲染行为与未知占位符保留的单测 |
CHANGELOG.md |
记录 Unreleased 新增：system_prompt playbook 插值变量支持 |
Review details
- Files reviewed: 3/3 changed files
- Comments generated: 3
- Review effort level: Lite
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
Comment on lines
+112
to
+135
| def _render_system_prompt( | ||
| template: str, | ||
| *, | ||
| agent_id: str, | ||
| agent_name: str, | ||
| work_dir: str, | ||
| model: str | None, | ||
| ) -> str: | ||
| """Render ``{playbook}`` variables into an agent system prompt. | ||
|
|
||
| Inspired by openresearch-cli's playbook substitution: the persona system | ||
| prompt may reference the agent's own identity, working directory, and the | ||
| current date so a single template stays accurate across instances. Values | ||
| are filled once when the harness graph is compiled, not per turn. | ||
| """ | ||
| now = datetime.datetime.now() | ||
| substitutions = { | ||
| "agent_id": agent_id, | ||
| "agent_name": agent_name, | ||
| "date": now.strftime("%Y-%m-%d"), | ||
| "datetime": now.strftime("%Y-%m-%d %H:%M"), | ||
| "work_dir": work_dir, | ||
| "model": model or "", | ||
| } |
Comment on lines
+2635
to
+2641
| system_prompt = _render_system_prompt( | ||
| system_prompt, | ||
| agent_id=row.agent_id, | ||
| agent_name=row.name or row.agent_id, | ||
| work_dir=str(harness_workspace), | ||
| model=default_model, | ||
| ) |
| assert "bot" in cfg.system_prompt # row.name | ||
| assert str(ws) in cfg.system_prompt | ||
| assert "gpt-4o-mini" in cfg.system_prompt | ||
| assert "20" in cfg.system_prompt # rendered date starts with 20xx |
…ndering
PersonaLoader.render used str.format, which raises KeyError on any
placeholder beyond {agent_name}/{user_display}/{custom}. Playbook
variables rendered later by AgentManager._render_system_prompt
({date}/{work_dir}/{model}/...) written into a custom persona would
crash agent creation. Switch to str.replace for the three known
placeholders so unknown ones pass through for compile-time rendering.
- persona.py: render uses replace; empty custom drops the placeholder line
- test_persona.py: unknown-placeholder preservation + empty-custom case
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
背景
Octop 的 system_prompt（agents 表 persona）是静态模板：同一个模板在多个 agent 实例间复用，但实例名/工作目录/时间无法区分。吸收 openresearch-cli 的 playbook 插值设计（SYSTEM_PROMPT.md 渲染 {name}/{id}/{project_state}），让 persona 模板支持身份与环境的占位符，编译图时渲染。
改动
AgentManager._render_system_prompt：渲染{agent_id}{agent_name}{date}{datetime}{work_dir}{model}六个变量；未知{xxx}原样保留（模板写错变量名不静默丢失）_build_harness_config：bootstrap 完成后非空 system_prompt 走渲染；default_model 只解析一次（渲染 + HarnessAgentConfig 复用）PersonaLoader.render：str.format改为宽容替换——用户自定义 persona 里的{date}等变量此前直接 KeyError 炸 agent 创建，现原样保留给编译期渲染设计边界
配套的上下文治理设计（本 PR 之外的本地补丁，供参考）
本 PR 是上下文治理的一部分，另外三块在 harness/deepagents 层（本机补丁，harness-agent 上游仓未公开，故不在此仓落地）：
测试计划