@delta-infra/cli 1.0.79 → 1.0.80

package/skills/antbo-ldm-guard/SKILL.md

@@ -1,58 +0,0 @@
----
-name: antbo-ldm-guard
-description: 任何同时涉及 AntBO/抗体/CDRH3 与 LDM/suggest/下一条候选序列的请求都必须优先于所有其他 AntBO Skill 立即使用本 Skill，即使历史序列或评分尚未提供。当前 Delta CLI catalog 未暴露 AntBO LDM initialize/evaluate/suggest；唯一正确结果是说明未发送远端请求，禁止询问数据、建文件、调用 health/Science/filesystem 或改用小分子 LDM-BO。
-metadata:
-  allowed-tools:
-    - python_repl
-    - step_finish
----
-
-# AntBO LDM 能力边界
-
-本 Skill 没有可执行的 Science operation，也没有 wrapper、endpoint 或 payload 示例。
-为了兼容 Kimi 对“无 artifact 的 step_finish”误判为 loop，先生成一个只记录能力边界的
-本地证据文件；这不是 Science 调用，也不发送远端请求。
-
-第一次工具调用必须是 `python_repl`，工具参数只能包含 `code`，执行以下等价逻辑：
-
-```python
-import json, os
-from pathlib import Path
-
-result = {
-    "supported": False,
-    "tool": "antbo",
-    "operation": "ldm-suggest",
-    "remote_request_sent": False,
-    "reason": "当前 Delta CLI catalog 未暴露 antbo LDM suggest operation",
-}
-output_path = Path(os.environ["WORKSPACE_ROOT"]) / "antbo_ldm_unsupported.json"
-output_path.write_text(json.dumps(result, ensure_ascii=False, indent=2), encoding="utf-8")
-print(str(output_path))
-```
-
-随后调用且只调用一次：
-
-```text
-step_finish(
-  status="done",
-  summary="当前 Delta CLI catalog 未暴露 antbo LDM suggest operation；未发送远端请求。",
-  output_ref=<刚生成的 antbo_ldm_unsupported.json 绝对路径>
-)
-```
-
-必须逐字使用上述 summary 和真实 output_ref。调用后立即结束，不输出其他文字，不调用
-任何工具。外层若核验，只能原子 `read_file(output_ref)` 一次，禁止其他操作。
-
-禁止行为：
-
-- 除上述一次本地文件写入外，禁止其他 `python_repl`；禁止 Bash、health、catalog、
-  filesystem、目录 read/list、ask_user；
-- 禁止创建 history JSON，禁止先要求用户提供序列或评分；
-- 禁止调用 `antbo-service`、`ldm-bo` 或通用子代理；
-- 禁止生成、转换、评价或声称返回候选 CDRH3；
-- 禁止把 AntBO health 公布的公司内部 endpoint 当成 Delta CLI operation。
-
-公司服务能力与 Delta CLI catalog 是两层能力。当前 Memento 只能通过 CLI catalog
-调用 `antbo` 的 `health/run-default-job/run/log/jobs/stop`；因此 AntBO LDM 请求必须在
-本 Skill 内以固定能力边界答复结束。

package/skills/antbo-ldm-guard/agents/openai.yaml

@@ -1,4 +0,0 @@
-interface:
-  display_name: "AntBO LDM 能力边界"
-  short_description: "识别当前 CLI 未暴露的 AntBO LDM 候选请求并安全终止"
-  default_prompt: "使用 $antbo-ldm-guard 判断这项 AntBO LDM 请求是否可执行。"

package/skills/antbo-service/SKILL.md

@@ -1,79 +0,0 @@
----
-name: antbo-service
-description: 当用户明确要求检查 AntBO 服务健康状态、启动默认或自定义 AntBO 作业、读取 AntBO 日志、列出活动作业或停止指定作业时使用。当前 Delta CLI operation 为 health、run-default-job、run、log、jobs、stop。
-metadata:
-  allowed-tools:
-    - read_file
-    - python_repl
-    - step_finish
----
-
-# AntBO 服务
-
-## 路由排除
-
-AntBO/CDRH3 LDM、suggest、下一条候选序列请求不进入本 Skill，必须由
-`antbo-ldm-guard` 处理。不得先调用本 Skill 的 health。
-
-## 路由与能力边界
-
-当前 Delta CLI catalog 暴露：
-
-- `health`：健康检查，只读。
-- `run-default-job`：启动默认后台任务，有副作用。
-- `run`：自定义同步任务，有副作用且可能长时间运行。
-- `log`：读取日志，只读。
-- `jobs`：列出当前容器登记的活动任务，只读。
-- `stop`：停止任务，有副作用。
-
-当前没有 `ldm/suggest`、`ldm/initialize` 或 `ldm/evaluate`。这些请求严格执行最上方
-终止分支；health 返回的公司内部 endpoint 不能覆盖 CLI catalog。
-
-## 强制执行规则
-
-1. 实时 Science 调用的第一个执行工具只能是 `python_repl`。
-   `python_repl` 工具参数只能包含 `code`，禁止 `deps/dependencies/packages`。
-2. 只通过 `os.environ["SKILLS_ROOT"]` 定位
-   `delta-science/scripts/invoke.py`，使用 `subprocess.run(..., shell=False)`。
-3. 禁止 Bash、`list_dir`、`read_file`、`which`、`where`、`ls`、wrapper/环境检查，
-   禁止执行字面命令 `CLI`、`cli` 或 `delta-cli`。
-4. 禁止 `curl`、`requests`、`httpx`、浏览器、网关直连和本地 AntBO。
-5. 业务结果直接位于 `result["native"]`；禁止读取 `native.data`。
-6. 每个 operation 的场景、最小参数和投影必须按
-   `references/operations.md`，不得猜相近 endpoint。
-
-## 唯一 wrapper 模板
-
-```python
-import json, os, subprocess, sys
-wrapper = os.path.join(os.environ["SKILLS_ROOT"], "delta-science", "scripts", "invoke.py")
-argv = [sys.executable, wrapper, "--tool", "antbo", "--endpoint", endpoint]
-if data is not None:
-    argv += ["--data-json", json.dumps(data)]
-if params is not None:
-    argv += ["--params-json", json.dumps(params)]
-p = subprocess.run(argv, capture_output=True, text=True, shell=False)
-if p.returncode != 0:
-    raise RuntimeError(p.stderr or p.stdout)
-r = json.loads(p.stdout)
-if r.get("ok") is not True or r.get("transport") != "delta-cli":
-    raise RuntimeError(r)
-n = r["native"]
-print(json.dumps({"transport": r["transport"], "endpoint": r["endpoint"], "native": n}, ensure_ascii=False))
-```
-
-执行前只按 `references/operations.md` 设置 `endpoint`、`data`、`params`。成功后将
-结果投影为用户要求的最小字段，不打印完整日志或 envelope。
-
-## 副作用与完成回执
-
-- `run-default-job`、`run`、`stop` 只有用户明确授权时才能执行。
-- 变更操作结果未知、超时或断连时禁止重试；客户端失败不表示远程操作未发生。
-- 只有同一次 `run-default-job` native 同时含 `started=true`、非空 `pid` 以及
-  `log_name` 或 `log_path` 时，才能说任务已启动。
-- `run` 仅用于用户明确提供并授权的受限脚本/config/antigen 路径，不生成任意路径。
-- `stop` 优先使用用户确认的 PID；不得在多任务不明确时无参数停止。
-
-成功后立即调用 `step_finish(status="done", summary=<有界结果 JSON>)` 恰好一次；
-失败时调用 `step_finish(status="failed", summary=<原始错误短摘要>)`。调用后不得输出
-自由文本或使用其他工具。

package/skills/antbo-service/agents/openai.yaml

@@ -1,4 +0,0 @@
-interface:
-  display_name: "AntBO 抗体优化服务"
-  short_description: "通过 Delta CLI 启动、查询或停止 AntBO 优化作业"
-  default_prompt: "使用 $antbo-service 执行这项 AntBO 作业管理任务。"

package/skills/delta-bo/SKILL.md

@@ -1,36 +0,0 @@
----
-name: delta-bo
-description: 当用户需要列出当前 Delta-BO 命令，或针对数值、整数、分类实验变量生成随机搜索/贝叶斯优化建议时使用。当前 Delta CLI catalog 只暴露 commands 与 generate；不得通过 Skill 猜测或直连旧的 stateful command API。
-metadata:
-  allowed-tools:
-    - read_file
-    - python_repl
-    - step_finish
----
-
-# Delta-BO 服务
-
-## 强制规则
-
-- 只用 `python_repl` 执行 `$SKILLS_ROOT/delta-science/scripts/invoke.py`，
-  tool=`delta-bo`；禁止 Bash、路径探测、字面 CLI、直接 HTTP 和本地 BO。
-- `python_repl` 工具参数只能包含 `code`，禁止 `deps/dependencies/packages`。
-- 业务结果直接在 `r["native"]`，禁止 `native.data`。
-- 当前 catalog 只有 `commands`（无 body）与 `generate`（JSON body）。旧文档中的
-  session/space/objective/policy/suggest/observe/loop 等不是当前 CLI operation，
-  不得执行或猜 endpoint。
-
-## generate 最小规则
-
-- 没有真实历史，或历史少于 5 条且用户未明确要求 BO：用
-  `algorithm:"random-search"`，给有界 `params`、`num_suggestions` 和 seed。
-- 有真实历史且明确要求 BO：使用 `objectives`、object-form
-  `histories.decisions/observations`，必要时才加入 model/acquisition/solver。
-- 参数类型仅 `numeric`、`integer`、`categorical`；不得编造观测或最优结果。
-- 只报告 native 的 suggestions、metadata、warnings、diversity；空建议或失败不得用
-  本地 GP/UCB/随机数替代。
-
-最小 payload 见 `references/endpoints.md` 的 Stateless Generate。通过 wrapper 一次
-调用、校验并显式打印有界 JSON，然后立即
-`step_finish(status="done", summary=<同一 JSON>)`；失败则 `status="failed"`。
-调用后不得输出自由文本或使用其他工具。

package/skills/delta-bo/agents/openai.yaml

@@ -1,4 +0,0 @@
-interface:
-  display_name: "Delta-BO 实验优化"
-  short_description: "通过 Delta CLI 为数值、整数和分类变量生成实验建议"
-  default_prompt: "使用 $delta-bo 根据变量空间生成下一批实验建议。"

package/skills/delta-bo/references/endpoints.md

@@ -1,263 +0,0 @@
-# Delta-BO API Endpoint Reference
-
-All paths are relative to `http://111.2.199.31:52317/api/v1`.
-
-This reference is for using the exposed HTTP API. It is not a demo runner.
-
-## Health And Command Discovery
-
-`GET /chem/delta-bo/health`
-
-`GET /chem/delta-bo/commands`
-
-`/commands` returns the native route list exposed by the deployed service.
-
-For a formatted catalog with gateway URLs against the deployed service, run the
-bundled skill-local script (no local `delta_bo` package required):
-
-```bash
-python3 scripts/remote_list_eng_cmds.py --format markdown
-python3 scripts/remote_list_eng_cmds.py --format json
-```
-
-For complete local route metadata including parameters, the offline equivalent
-`list_eng_cmds.py` (requires a Delta-BO dev environment with dependencies
-installed, not bundled) walks the same command registry used by the service
-dispatcher and is the source of truth for possible native API usage.
-
-## Stateless Generate
-
-`POST /chem/delta-bo/generate`
-
-Random search:
-
-```json
-{
-  "params": [{"name": "x", "type": "numeric", "low": 0.0, "high": 1.0}],
-  "num_suggestions": 3,
-  "algorithm": "random-search",
-  "seed": 42
-}
-```
-
-BO with history:
-
-```json
-{
-  "params": [
-    {"name": "temperature", "type": "numeric", "low": 20.0, "high": 120.0},
-    {"name": "catalyst", "type": "categorical", "choices": ["A", "B", "C"]}
-  ],
-  "objectives": [{"name": "yield", "minimize": false}],
-  "num_suggestions": 2,
-  "algorithm": "bo",
-  "model": {"name": "gp", "min_samples": 2, "train_steps": 50},
-  "acq_target": {"name": "ucb", "beta": 2.0},
-  "solver": {"name": "ga", "pop_size": 40, "n_gen": 10},
-  "histories": {
-    "decisions": [[40.0, "A"], [80.0, "B"], [100.0, "C"]],
-    "observations": [0.42, 0.61, 0.53]
-  },
-  "seed": 7
-}
-```
-
-Parameter types: `numeric`, `integer`, `categorical`.
-
-Histories may also use object-form rows, which are safer for agents:
-
-```json
-{
-  "histories": {
-    "decisions": [
-      {"temperature": 40.0, "catalyst": "A"},
-      {"temperature": 80.0, "catalyst": "B"}
-    ],
-    "observations": [
-      {"yield": 0.42},
-      {"yield": 0.61}
-    ]
-  }
-}
-```
-
-Optional diversity controls:
-
-```json
-{
-  "diversity": {
-    "deduplicate": true,
-    "min_numeric_distance": {"temperature": 5.0},
-    "prefer_unique_categories": true
-  },
-  "constraints": [
-    {"type": "unique_batch"},
-    {"type": "avoid_seen_decisions"},
-    {
-      "type": "max_temperature_for_category",
-      "category_param": "solvent",
-      "temperature_param": "temperature",
-      "category": "dichloromethane",
-      "max": 40
-    }
-  ]
-}
-```
-
-Response fields include:
-
-```json
-{
-  "suggestions": [[{"name": "temperature", "value": 72.0}]],
-  "suggestion_metadata": [
-    {
-      "acquisition_value": null,
-      "nearest_history": {
-        "decision": [{"name": "temperature", "value": 80.0}],
-        "distance": 0.08
-      },
-      "rationale": "exploitation"
-    }
-  ],
-  "warnings": [
-    "Only 3 observations supplied; GP fit may be weak. Consider random-search or more initial points."
-  ],
-  "generated_candidate_count": 15,
-  "diversity": {
-    "requested": true,
-    "deduplicated": true,
-    "duplicates_removed": 2
-  }
-}
-```
-
-## Native Stateful Command API
-
-Gateway path pattern:
-
-```text
-POST /chem/delta-bo/v1/{command_path}
-```
-
-Native route `/v1/session/new` maps to gateway route:
-
-```text
-/api/v1/chem/delta-bo/v1/session/new
-```
-
-Minimal random-search session:
-
-```bash
-curl -s "$DBO/v1/session/new" \
-  -H "Content-Type: application/json" \
-  -d '{"name":"api_session"}'
-
-curl -s "$DBO/v1/space" \
-  -H "Content-Type: application/json" \
-  -d '{"decision_expression":"Joint({\"x\": Num(-5, 10), \"y\": Num(0, 15)})"}'
-
-curl -s "$DBO/v1/objective" \
-  -H "Content-Type: application/json" \
-  -d '{"source":"objectives/branin_simple.py","minimize":true}'
-
-curl -s "$DBO/v1/policy/random-search" \
-  -H "Content-Type: application/json" \
-  -d '{"policy_id":"random_init","replace":true}'
-
-curl -s "$DBO/v1/suggest" \
-  -H "Content-Type: application/json" \
-  -d '{"num":3,"policy_id":"random_init"}'
-```
-
-Common native commands:
-
-```text
-session/new
-session/save
-space
-objective
-policy/random-search
-policy/bo
-policy/bo.model/gp
-policy/bo.target/ucb
-policy/bo.target/ei
-policy/bo.solver/ga
-policy/flags
-suggest
-observe
-loop
-status
-best
-plot/best-so-far
-```
-
-Use `GET /chem/delta-bo/commands` for the deployed list; run the bundled
-`scripts/remote_list_eng_cmds.py --format markdown/json` for a formatted catalog
-against the deployed service, or the offline `list_eng_cmds.py` (local Delta-BO
-dev environment, not bundled) for full parameter schemas.
-
-## Native BO Policy Setup
-
-A typical stateful BO configuration is:
-
-```bash
-curl -s "$DBO/v1/policy/bo" \
-  -H "Content-Type: application/json" \
-  -d '{"replace":true}'
-
-curl -s "$DBO/v1/policy/bo.model/gp" \
-  -H "Content-Type: application/json" \
-  -d '{"train_steps":50,"min_samples":2}'
-
-curl -s "$DBO/v1/policy/bo.target/ucb" \
-  -H "Content-Type: application/json" \
-  -d '{"beta":2.0}'
-
-curl -s "$DBO/v1/policy/bo.solver/ga" \
-  -H "Content-Type: application/json" \
-  -d '{"pop_size":40,"n_gen":10}'
-```
-
-For random initialization policy creation:
-
-```bash
-curl -s "$DBO/v1/policy/random-search" \
-  -H "Content-Type: application/json" \
-  -d '{"policy_id":"random_init","replace":true}'
-```
-
-Then run an optimization loop:
-
-```bash
-curl -s "$DBO/v1/loop" \
-  -H "Content-Type: application/json" \
-  -d '{"iterations":5,"batch_size":1,"policy_id":"main","timeout":300}'
-```
-
-Response shape is nested:
-
-```json
-{
-  "code": 0,
-  "message": "success",
-  "data": {
-    "ok": true,
-    "result": {
-      "data": {},
-      "summary": "...",
-      "meta": {"command_path": "/v1/..."}
-    }
-  }
-}
-```
-
-## Retry Safety
-
-Setup/configuration requests such as `space`, `objective`, `policy/bo`, and
-`policy/random-search` can be repeated with the same payload. Use
-`{"replace": true}` for policy creation so repeated setup calls remain
-idempotent.
-
-Do not blindly retry state-changing optimization commands such as `loop`,
-`suggest`, `observe`, `session/new`, or `session/save`, because a retry may
-advance or duplicate remote experiment state.

package/skills/delta-science/references/generated-catalog.md

@@ -1,48 +0,0 @@
-# Science catalog（自动生成）
-
-本文件由 `skills/science-tools.json` 生成；不要手工修改。规范名称供 CLI/Skill 使用，legacy 名称仅用于兼容已部署的 `/science_tool` catalog。
-
-| 规范 tool | Skill | 类别 | operation | legacy endpoint |
-| --- | --- | --- | --- | --- |
-| pubchem | pubchem | molecule-information | health | chem_pubchem_health |
-| pubchem | pubchem | molecule-information | compound-resolve-cids | chem_pubchem_compound_resolve-cids |
-| pubchem | pubchem | molecule-information | compound-properties | chem_pubchem_compound_properties |
-| pubchem | pubchem | molecule-information | compound-synonyms | chem_pubchem_compound_synonyms |
-| pubchem | pubchem | molecule-information | compound-summary | chem_pubchem_compound_summary |
-| pubchem | pubchem | molecule-information | compound-batch-summary | chem_pubchem_compound_batch-summary |
-| rdkit | rdkit | cheminformatics | health | chem_rdkit_health |
-| rdkit | rdkit | cheminformatics | parse | chem_rdkit_parse |
-| rdkit | rdkit | cheminformatics | descriptors | chem_rdkit_descriptors |
-| rdkit | rdkit | cheminformatics | batch-descriptors | chem_rdkit_batch-descriptors |
-| rdkit | rdkit | cheminformatics | batch-parse-describe | chem_rdkit_batch-parse-describe |
-| rdkit | rdkit | cheminformatics | render | chem_rdkit_render |
-| rdkit | rdkit | cheminformatics | fingerprint | chem_rdkit_fingerprint |
-| rdkit | rdkit | cheminformatics | similarity | chem_rdkit_similarity |
-| rdkit | rdkit | cheminformatics | similarity-matrix | chem_rdkit_similarity-matrix |
-| rdkit | rdkit | cheminformatics | substructure | chem_rdkit_substructure |
-| pymatgen | pymatgen | materials-informatics | health | chem_pymatgen_health |
-| pymatgen | pymatgen | materials-informatics | composition-parse | chem_pymatgen_composition_parse |
-| pymatgen | pymatgen | materials-informatics | structure-parse | chem_pymatgen_structure_parse |
-| pymatgen | pymatgen | materials-informatics | structure-summary | chem_pymatgen_structure_summary |
-| pymatgen | pymatgen | materials-informatics | structure-convert | chem_pymatgen_structure_convert |
-| pymatgen | pymatgen | materials-informatics | structure-symmetry | chem_pymatgen_structure_symmetry |
-| gsasii | gsasii | diffraction | health | chem_gsasii_health |
-| gsasii | gsasii | diffraction | powder-simulate | chem_gsasii_powder_simulate |
-| gsasii | gsasii | diffraction | powder-refine | chem_gsasii_powder_refine |
-| lammps | lammps | molecular-dynamics | health | chem_lammps_health |
-| lammps | lammps | molecular-dynamics | lj-melt-example | chem_lammps_examples_lj-melt |
-| lammps | lammps | molecular-dynamics | run | chem_lammps_run |
-| delta-bo | delta-bo | optimization | commands | chem_delta-bo_commands |
-| delta-bo | delta-bo | optimization | generate | chem_delta-bo_generate |
-| ldm-bo | ldm-bo | molecule-optimization | health | chem_ldm-bo_health |
-| ldm-bo | ldm-bo | molecule-optimization | recommend | chem_ldm-bo_recommend |
-| ldm-bo | ldm-bo | molecule-optimization | trajectory | chem_ldm-bo_trajectory |
-| synbo | synbo-service | reaction-optimization | health | chem_synbo_health |
-| synbo | synbo-service | reaction-optimization | initialize | chem_synbo_initialize |
-| synbo | synbo-service | reaction-optimization | optimize | chem_synbo_optimize |
-| antbo | antbo-service | antibody-optimization | health | biology_antbo_health |
-| antbo | antbo-service | antibody-optimization | run-default-job | biology_antbo_run_default_job |
-| antbo | antbo-service | antibody-optimization | run | biology_antbo_run |
-| antbo | antbo-service | antibody-optimization | log | biology_antbo_log |
-| antbo | antbo-service | antibody-optimization | jobs | biology_antbo_jobs |
-| antbo | antbo-service | antibody-optimization | stop | biology_antbo_stop |

package/skills/delta-science/references/operation-contracts.md

@@ -1,137 +0,0 @@
-# Delta Science operation 最小契约
-
-本文只记录当前 CLI catalog 已暴露的 9 个 tools、42 个 operations。选择 operation 后
-只读取对应小节。代码示例均在同一个 `python_repl` 中先定义以下函数；禁止 Bash、
-目录探测、直接 HTTP 和字面 CLI 命令。
-
-调用 `python_repl` 时工具参数必须严格为 `{"code":"..."}`；禁止添加 `deps`、
-`dependencies`、`packages` 或任何其他字段。
-
-```python
-import json, os, subprocess, sys
-
-wrapper = os.path.join(os.environ["SKILLS_ROOT"], "delta-science", "scripts", "invoke.py")
-
-def invoke(tool, endpoint, data=None, params=None):
-    argv = [sys.executable, wrapper, "--tool", tool, "--endpoint", endpoint]
-    if data is not None:
-        argv += ["--data-json", json.dumps(data)]
-    if params is not None:
-        argv += ["--params-json", json.dumps(params)]
-    p = subprocess.run(argv, capture_output=True, text=True, shell=False)
-    if p.returncode != 0:
-        raise RuntimeError(p.stderr or p.stdout)
-    r = json.loads(p.stdout)
-    if r.get("ok") is not True or r.get("transport") != "delta-cli":
-        raise RuntimeError(r)
-    return r["native"]
-```
-
-每个代码示例只调用一次。将返回值投影成用户要求的有界 JSON 并显式 `print`，然后
-调用 `step_finish(status="done", summary=<同一 JSON>)` 恰好一次。失败则
-`step_finish(status="failed", summary=<原始错误短摘要>)`，禁止重试或改走其他链路。
-
-## PubChem（tool=`pubchem`）
-
-| operation | 人类情景 | 最小代码 | native 投影 |
-| --- | --- | --- | --- |
-| `health` | “PubChem 服务现在能用吗？” | `n=invoke("pubchem","health")` | status、availability、metadata、limits、warnings |
-| `compound-resolve-cids` | “把 aspirin 解析成 CID。” | `n=invoke("pubchem","compound-resolve-cids",data={"identifier":"aspirin","namespace":"name","max_records":20})` | identifier、valid、cids、error |
-| `compound-properties` | “查 aspirin 的分子式和分子量。” | `n=invoke("pubchem","compound-properties",data={"identifier":"aspirin","namespace":"name","properties":["MolecularFormula","MolecularWeight"]})` | cid、valid、properties、aliases、error |
-| `compound-synonyms` | “列出 aspirin 的前 5 个同义词。” | `n=invoke("pubchem","compound-synonyms",data={"identifier":"aspirin","namespace":"name","max_synonyms":5})` | cid、valid、synonyms、error |
-| `compound-summary` | “查阿司匹林的基础身份信息。” | `n=invoke("pubchem","compound-summary",data={"identifier":"aspirin","namespace":"name"})` | valid、cid、properties 中用户要求的键 |
-| `compound-batch-summary` | “一次比较 aspirin 和 caffeine 的基础属性。” | `n=invoke("pubchem","compound-batch-summary",data={"identifiers":["aspirin","caffeine"],"namespace":"name"})` | results[*].input、valid/error、properties |
-
-## RDKit（tool=`rdkit`）
-
-| operation | 人类情景 | 最小代码 | native 投影 |
-| --- | --- | --- | --- |
-| `health` | “RDKit 服务可用吗？” | `n=invoke("rdkit","health")` | status、availability、metadata、limits、warnings |
-| `parse` | “这个 SMILES 合法吗？” | `n=invoke("rdkit","parse",data={"input":"CCO","format":"smiles","sanitize":True})` | valid、canonical_smiles、formula、inchi_key、warnings/error |
-| `descriptors` | “计算乙醇的基础 RDKit 描述符。” | `n=invoke("rdkit","descriptors",data={"smiles":"CCO","descriptor_set":"basic","sanitize":True})` | valid、descriptors、warnings/error |
-| `batch-descriptors` | “批量算几个 SMILES 的描述符。” | `n=invoke("rdkit","batch-descriptors",data={"smiles":["CCO","c1ccccc1"],"descriptor_set":"basic","sanitize":True})` | results 中输入、valid、descriptors、error |
-| `batch-parse-describe` | “校验阿司匹林并给基础描述符。” | `n=invoke("rdkit","batch-parse-describe",data={"molecules":[{"id":"aspirin","smiles":"CC(=O)OC1=CC=CC=C1C(=O)O"}],"descriptor_set":"basic","sanitize":True})` | results[*].id、valid、canonical_smiles、descriptors、warnings/error |
-| `render` | “画一张乙醇结构图。” | `n=invoke("rdkit","render",data={"smiles":"CCO","image_format":"png","width":500,"height":400})` | 指定图像字段；PNG 解码后验证 magic bytes |
-| `fingerprint` | “生成乙醇 Morgan 指纹。” | `n=invoke("rdkit","fingerprint",data={"smiles":"CCO","radius":2,"n_bits":2048})` | valid、fingerprint_metadata、on_bits、warnings/error |
-| `similarity` | “比较乙醇和乙胺的结构相似度。” | `n=invoke("rdkit","similarity",data={"query_smiles":"CCO","target_smiles":"CCN","radius":2,"n_bits":2048})` | similarity、metric、fingerprint metadata |
-| `similarity-matrix` | “比较一组分子的两两相似度。” | `n=invoke("rdkit","similarity-matrix",data={"molecules":[{"id":"a","smiles":"CCO"},{"id":"b","smiles":"CCN"}],"fingerprint":{"type":"morgan","radius":2,"n_bits":2048},"sanitize":True})` | matrix、ranked_pairs[{a,b,similarity}] |
-| `substructure` | “筛出含羧酸子结构的分子。” | `n=invoke("rdkit","substructure",data={"query_smarts":"C(=O)O","target_smiles":["CC(=O)O","CCO"]})` | query 与每个目标的 match/matches、valid/error |
-
-## pymatgen（tool=`pymatgen`）
-
-| operation | 人类情景 | 最小代码 | native 投影 |
-| --- | --- | --- | --- |
-| `health` | “pymatgen 服务是否正常？” | `n=invoke("pymatgen","health")` | status、availability、metadata、limits、warnings |
-| `composition-parse` | “解析 LiFePO4 的组成和式量。” | `n=invoke("pymatgen","composition-parse",data={"formula":"LiFePO4","format":"formula"})` | valid、formula、reduced_formula、chemical_system、weight、element_amounts |
-| `structure-parse` | “校验这段 CIF 能否解析。” | `n=invoke("pymatgen","structure-parse",data={"input":cif_text,"format":"cif","primitive":False})` | valid、formula、nsites、lattice、warnings/error |
-| `structure-summary` | “总结这段 CIF 的结构信息。” | `n=invoke("pymatgen","structure-summary",data={"input":cif_text,"format":"cif","primitive":False})` | formula、reduced_formula、density、volume、lattice、species |
-| `structure-convert` | “把 CIF 转成 POSCAR。” | `n=invoke("pymatgen","structure-convert",data={"input":cif_text,"input_format":"cif","output_format":"poscar","primitive":False})` | valid、output_format、转换后的结构文本、warnings/error |
-| `structure-symmetry` | “分析这段 CIF 的空间群。” | `n=invoke("pymatgen","structure-symmetry",data={"input":cif_text,"format":"cif","primitive":False,"symprec":0.01,"angle_tolerance":5.0})` | space_group_symbol/number、crystal_system、point_group_symbol、hall、nsites |
-
-## GSAS-II（tool=`gsasii`）
-
-| operation | 人类情景 | 最小代码 | native 投影 |
-| --- | --- | --- | --- |
-| `health` | “GSAS-II 服务能做衍射计算吗？” | `n=invoke("gsasii","health")` | status、gsasii_available、metadata、limits、warnings |
-| `powder-simulate` | “用这个 CIF 模拟 Cu Kα 粉末衍射。” | `n=invoke("gsasii","powder-simulate",data={"cif":cif_text,"phase_name":"phase","radiation":"CuKa","t_min":10.0,"t_max":80.0,"t_step":0.02,"include_profile":True,"max_profile_points":1000,"include_reflections":True,"max_reflections":1000,"return_refined_cif":False,"return_project":False})` | valid、profile、reflections、metadata、warnings |
-| `powder-refine` | “用真实粉末数据和仪器文件做 Rietveld 精修。” | `n=invoke("gsasii","powder-refine",data={"powder_data":powder_text,"powder_format":"xy","instrument_parameters":instrument_text,"phases":[{"name":"phase","cif":cif_text}],"refinement_mode":"rietveld","refinement_steps":[],"max_cycles":3,"histogram_name":"powder","include_profile":True,"max_profile_points":1000,"include_reflections":True,"max_reflections":1000,"return_refined_cif":False,"return_project":False})` | valid、residuals、phase summaries、profile、reflections、warnings |
-
-## LAMMPS（tool=`lammps`）
-
-| operation | 人类情景 | 最小代码 | native 投影 |
-| --- | --- | --- | --- |
-| `health` | “LAMMPS 服务可用吗？” | `n=invoke("lammps","health")` | status、lammps_available、metadata、limits、warnings |
-| `lj-melt-example` | “给我跑部署自带的最小 LJ 示例做检查。” | `n=invoke("lammps","lj-melt-example")` | 示例说明与精确 `request`；此调用本身不执行模拟 |
-| `run` | “执行刚才返回的最小 LJ 请求。” | `example=invoke("lammps","lj-melt-example"); n=invoke("lammps","run",data=example["request"])` | valid、last_thermo、thermo rows、warnings、请求的输出文件；此场景固定两次依赖调用 |
-
-`run` 也可接收用户明确提供的 `input_script`、相对路径 files、timeout 和有界输出配置；
-不得从化学式猜力场。长任务需明确授权，失败后不得重试。
-
-## Delta-BO（tool=`delta-bo`）
-
-| operation | 人类情景 | 最小代码 | native 投影 |
-| --- | --- | --- | --- |
-| `commands` | “当前 Delta-BO 支持哪些命令？” | `n=invoke("delta-bo","commands")` | 当前部署返回的命令列表，不从旧文档补充 |
-| `generate` | “在 0 到 1 之间给两个随机初始实验点。” | `n=invoke("delta-bo","generate",data={"params":[{"name":"x","type":"numeric","low":0.0,"high":1.0}],"num_suggestions":2,"algorithm":"random-search","seed":42})` | suggestions、suggestion_metadata、warnings、diversity |
-
-有历史观测且用户明确要求 BO 时，`generate` 可加入 objectives、object-form histories、
-model/acquisition/solver；历史少于 5 条默认 random-search，不编造观测。
-
-## LDM-BO（tool=`ldm-bo`）
-
-只用于小分子 SMILES/PDF2Dock，不用于抗体、CDRH3 或 AntBO。
-
-| operation | 人类情景 | 最小代码 | native 投影 |
-| --- | --- | --- | --- |
-| `health` | “分子 LDM-BO 服务可用吗？” | `n=invoke("ldm-bo","health")` | availability、Vina/ReaSyn/模型/LLM 配置状态、warnings |
-| `recommend` | “从候选池给下一条小分子 SMILES。” | `n=invoke("ldm-bo","recommend",data={"method":"bo-tanimoto","pool":["CCO","CCN","CCC"],"history":[{"smiles":"CCO","scores":[-7.1]},{"smiles":"CCN","scores":[-6.8]}],"batch_size":1})` | recommendations、acquisition_values、n_objectives、llm diagnostics |
-| `trajectory` | “跑一个有界的小分子 BO 轨迹。” | `n=invoke("ldm-bo","trajectory",data={"method":"bo-tanimoto","seed":42,"seed-smiles":"CCO,CCN,CCC","num-evaluations":4,"batch-size":1,"objective":"vina+nn"})` | config、history、summary、llm_trajectory；保留连字符字段 |
-
-## SynBO（tool=`synbo`）
-
-| operation | 人类情景 | 最小代码 | native 投影 |
-| --- | --- | --- | --- |
-| `health` | “SynBO 服务当前可用吗？” | `n=invoke("synbo","health")` | status、synbo_available、metadata、limits、runtime、warnings |
-| `initialize` | “没有实验历史，先选三组反应条件。” | `n=invoke("synbo","initialize",data={"condition_dict":{"catalyst":["Pd(OAc)2","Pd(PPh3)4"],"solvent":["THF","Toluene"],"base":["K2CO3","DBU"]},"opt_metrics":["yield"],"opt_metric_settings":[{"opt_direct":"max","opt_range":[0,100],"metric_weight":1.0}],"batch_size":3,"sampling_method":"random","desc_normalize":"minmax","random_seed":42})` | recommendations/initial conditions、warnings、metadata |
-| `optimize` | “根据真实历史收率推荐下一批条件。” | `n=invoke("synbo","optimize",data={"condition_dict":{"catalyst":["Pd(OAc)2","Pd(PPh3)4","Pd2(dba)3"],"solvent":["THF","Dioxane","Toluene"],"base":["K2CO3","NaOEt","DBU"]},"opt_metrics":["yield"],"previous_results":[{"batch":0,"catalyst":"Pd(OAc)2","solvent":"THF","base":"K2CO3","yield":42.0},{"batch":0,"catalyst":"Pd(PPh3)4","solvent":"Dioxane","base":"NaOEt","yield":55.0},{"batch":0,"catalyst":"Pd2(dba)3","solvent":"Toluene","base":"DBU","yield":63.0}],"batch_size":2,"accuracy":"tiny","device":"cpu","random_seed":42})` | recommendations、model/acquisition metadata、warnings；失败/空结果不得本地替代 |
-
-## AntBO（tool=`antbo`）
-
-抗体/CDRH3 场景即使出现 “LDM” 也必须路由到 AntBO，不得改用分子 LDM-BO。
-
-| operation | 人类情景 | 最小代码 | native 投影 |
-| --- | --- | --- | --- |
-| `health` | “AntBO 环境可用吗？” | `n=invoke("antbo","health")` | status、antbo_available、metadata、limits、warnings |
-| `run-default-job` | “启动标准 AntBO 后台任务。” | `n=invoke("antbo","run-default-job",params={"log_name":"antbo_<唯一时间戳>.log","append":False,"timeout_seconds":86400})` | started、pid、log_name/log_path、timeout_seconds |
-| `run` | “用明确的受限配置同步跑一次 AntBO。” | `n=invoke("antbo","run",data={"script_path":"./bo/main.py","config_path":"./bo/config.yaml","n_trials":1,"seed":42,"antigens_file":"<项目根内路径>","extra_args":[],"timeout_seconds":86400,"cuda_visible_devices":"0","conda_env":"DGM","log_name":"antbo_custom.log","append":False})` | ok、returncode、timed_out、log_path、log_truncated |
-| `log` | “查看 antbo_run.log 的尾部。” | `n=invoke("antbo","log",params={"log_name":"antbo_run.log"})` | log_name、log、log_truncated |
-| `jobs` | “查看当前活动 AntBO 作业。” | `n=invoke("antbo","jobs")` | jobs[*].pid/running/returncode/log_name/log_path |
-| `stop` | “停止 PID 12345。” | `n=invoke("antbo","stop",params={"pid":12345})` | stopped、pid、message |
-
-`run-default-job`、`run`、`stop` 是远程变更，必须有用户明确授权，未知结果不得重试。
-AntBO `ldm/suggest`、`ldm/initialize`、`ldm/evaluate` 当前不在 catalog：不调用工具，
-固定报告“未暴露该 operation；未发送远端请求”。
-
-注意：AntBO `health` 可能公布公司服务内部的 `ldm_suggest` 等 endpoint，但它们并不
-因此成为 Delta CLI operation；禁止调用不存在的 `antbo/catalog`，也禁止根据 health
-为 Memento 推导新的 endpoint。