notebook-dossier
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| NLM_ACTOR | No | Caller identifier written to the audit log, useful when multiple hosts share the same database. | pid:<pid>:<token> |
| NLM_APP_DIR | No | Directory for the task database, research artifacts, and exports. Defaults to the system user data directory for notebook-dossier. | system user data directory notebook-dossier |
| NLM_PROFILE | No | The upstream login profile to use, such as 'research'. Defaults to the upstream saved default profile. | upstream default profile |
| NLM_QUERY_TIMEOUT | No | Timeout in seconds for a single query. | 240 |
| NLM_UPSTREAM_ARGS | No | Arguments for the upstream command, separated by spaces. | |
| NLM_IMPORT_TIMEOUT | No | Timeout in seconds for import operations. | 300 |
| NLM_UPSTREAM_COMMAND | No | The upstream MCP command. Must be an absolute path when the child process PATH may not resolve it. | notebooklm-mcp |
| NLM_TASK_LEASE_SECONDS | No | Task lease duration in seconds. Another process may only take over a task after its heartbeat is older than this value. | 120 |
| NLM_SOURCE_WAIT_TIMEOUT | No | Timeout in seconds for waiting on source indexing. | 120 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
| logging | {} |
| prompts | {
"listChanged": true
} |
| resources | {
"subscribe": false,
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| dossier_auth_statusB | 查询 NotebookLM 上游的本地登录与在线状态（不返回任何凭据）。 |
| dossier_startA | 启动一次 NotebookLM Deep Research（web + deep），立即返回 research_id。 每个独立主题一个研究 Notebook。request_key 用于幂等：同键重复调用返回已有任务，不会重复发起研究。 |
| dossier_statusA | 查询研究任务状态，并同步完整报告与发现来源清单。 完成时会保存 compact=False 的完整结果；来源统计与发现/导入不是同一指标。 返回 phase/elapsed_ms/sources_discovered 与自适应 suggested_next_poll_seconds。 |
| dossier_resultA | 分页读取研究结果。view: findings | report | digest | evidence | comparisons | tasks | answers。 findings 为完整发现来源（含 cite_map、citation_warnings、逐来源提取式摘要）； report 为报告全文分页（cursor=字符偏移、max_chars=窗口）； digest 为提取式执行摘要（≤1500 字 + 高被引来源清单）； include_report_text 仅作用于 findings 视图的报告内嵌（旧口径，建议改用 view='report'）。 |
| dossier_importC | 把选中的发现来源导入 NotebookLM，并逐个记录导入映射与不确定结果。 |
| dossier_source_getB | 取得单个来源的元数据与索引文本摘录（含内容版本哈希与取得时间）。 来源元数据缺失的字段保持 null/未知，不从模型补齐。 |
| dossier_queryA | 向 NotebookLM 提问；来源集合由软件强制限定（默认新会话，逐来源取证只传一个来源）。 返回回答与可核验引用；引文编号仅在本回答内有效（citation_key 形如 research:answer_seq#n）。 |
| dossier_evidenceB | 逐来源生成证据卡（每来源单独会话、只依据该来源），返回任务号。 证据卡含主张、依据类型、短引文、定位、适用条件、局限与验证状态；短引文会与来源原文比对。 |
| dossier_compareA | 按共同问题比较各来源观点，生成共识/差异/缺口（后台任务，返回任务号）。 软件强制规则：仅未提及≠反对；冲突结论必须有双方原文引文；转载合并计、报告不重复计证据。 statements 可传宿主 AI 已形成的比较草稿（HostStatementInput schema），与 NotebookLM 比较结果合并校验。 |
| dossier_exportB | 导出文章研究底稿目录：manifest / 报告 / sources / evidence / viewpoints / writing-brief。 导出物不含 Cookie、令牌或浏览器 Profile；每个重要判断关联来源与证据 ID。 |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 10 tools
Most tools target distinct actions: start/status/result/query/import/export/source_get are clearly separable. The fuzziest boundary is between the analysis-producing trio (dossier_compare, dossier_evidence, dossier_query) and the ability to read those same artifacts via dossier_result's views (evidence/comparisons/answers), which requires careful reading of which tool creates versus just retrieves them.
Strong, predictable snake_case with a uniform 'dossier_' prefix across all ten tools. Minor inconsistency in that several are noun-only state/artifact tools (dossier_status, dossier_result, dossier_auth_status) while others use verbs (dossier_start, dossier_query, dossier_export), but the pattern remains readable.
Ten tools is well within the ideal range and each maps to a distinct step in the research lifecycle: launch, poll, read, ask, compare, evidence, import, export, inspect source, check auth. No obvious redundancy or padding in the surface.
Covers the full lifecycle from dossier_start through status/result, downstream analysis (query/compare/evidence), source inspection, import, and export, with good idempotency hints (request_key). Gaps exist: no way to list or enumerate existing research tasks/notebooks, and no cancel/delete/retry operation, so an agent that loses a research_id has a dead end.