Spaces:
Running
Running
| swagger: '2.0' | |
| info: | |
| title: InfoRadar API | |
| version: "0.1" | |
| consumes: | |
| - application/json | |
| produces: | |
| - application/json | |
| basePath: /api | |
| paths: | |
| /analyze: | |
| post: | |
| tags: | |
| - all | |
| summary: analyze text | |
| operationId: server.analyze | |
| parameters: | |
| - in: body | |
| name: analyze_request | |
| description: request analyze | |
| schema: | |
| $ref: "#/definitions/AnalyzeRequest" | |
| responses: | |
| 200: | |
| description: | | |
| 成功时为信息密度分析结果。`stream=false`（默认）时响应体即本 schema； | |
| `stream=true` 时为 `text/event-stream`，最后一条 `data` 事件内为同形 JSON。 | |
| 业务错误（4xx/5xx）时响应体含 `success`/`message` 等，见各状态码说明，非本 schema。 | |
| schema: | |
| $ref: "#/definitions/AnalyzeResponse" | |
| 400: | |
| description: invalid request payload (missing text or model) | |
| 404: | |
| description: requested model is not registered | |
| 500: | |
| description: model inference failed | |
| 503: | |
| description: requested model could not be loaded | |
| /list_demos: | |
| get: | |
| tags: | |
| - all | |
| summary: list all available demo files and folders | |
| operationId: server.list_demos | |
| parameters: | |
| - in: query | |
| name: path | |
| type: string | |
| required: false | |
| description: path to list (empty string for root) | |
| responses: | |
| 200: | |
| description: return list of demo files and folders | |
| schema: | |
| type: object | |
| properties: | |
| path: | |
| type: string | |
| description: current path | |
| items: | |
| type: array | |
| items: | |
| type: object | |
| properties: | |
| type: | |
| type: string | |
| enum: [folder, file] | |
| description: item type | |
| name: | |
| type: string | |
| description: item name | |
| path: | |
| type: string | |
| description: full path | |
| file: | |
| type: string | |
| description: relative file path (only for files) | |
| /save_demo: | |
| post: | |
| tags: | |
| - all | |
| summary: save demo file | |
| operationId: server.save_demo | |
| parameters: | |
| - in: body | |
| name: save_request | |
| description: request to save demo | |
| schema: | |
| type: object | |
| properties: | |
| name: | |
| type: string | |
| description: demo name (will be used as filename) | |
| data: | |
| type: object | |
| description: AnalyzeResponse data to save | |
| path: | |
| type: string | |
| description: "target folder path (default: \"/\" for root directory)" | |
| default: "/" | |
| required: | |
| - name | |
| - data | |
| responses: | |
| 200: | |
| description: return save result | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| message: | |
| type: string | |
| file: | |
| type: string | |
| description: saved filename | |
| /delete_demo: | |
| post: | |
| tags: | |
| - all | |
| summary: delete demo file | |
| operationId: server.delete_demo | |
| parameters: | |
| - in: body | |
| name: delete_request | |
| description: request to delete demo | |
| schema: | |
| type: object | |
| properties: | |
| file: | |
| type: string | |
| description: demo filename (with .json extension) | |
| required: | |
| - file | |
| responses: | |
| 200: | |
| description: return delete result | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| message: | |
| type: string | |
| /move_demo: | |
| post: | |
| tags: | |
| - all | |
| summary: move demo file or folder | |
| operationId: server.move_demo | |
| parameters: | |
| - in: body | |
| name: move_request | |
| description: request to move demo or folder | |
| schema: | |
| type: object | |
| properties: | |
| file: | |
| type: string | |
| description: file path (for moving files) | |
| path: | |
| type: string | |
| description: folder path (for moving folders) | |
| target_path: | |
| type: string | |
| description: target folder path (empty string for root) | |
| required: | |
| - target_path | |
| responses: | |
| 200: | |
| description: return move result | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| message: | |
| type: string | |
| /rename_demo: | |
| post: | |
| tags: | |
| - all | |
| summary: rename demo file | |
| operationId: server.rename_demo | |
| parameters: | |
| - in: body | |
| name: rename_request | |
| description: request to rename demo file | |
| schema: | |
| type: object | |
| properties: | |
| file: | |
| type: string | |
| description: file path | |
| new_name: | |
| type: string | |
| description: new name (without extension) | |
| required: | |
| - file | |
| - new_name | |
| responses: | |
| 200: | |
| description: return rename result | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| message: | |
| type: string | |
| /rename_folder: | |
| post: | |
| tags: | |
| - all | |
| summary: rename folder | |
| operationId: server.rename_folder | |
| parameters: | |
| - in: body | |
| name: rename_request | |
| description: request to rename folder | |
| schema: | |
| type: object | |
| properties: | |
| path: | |
| type: string | |
| description: folder path | |
| new_name: | |
| type: string | |
| description: new folder name | |
| required: | |
| - path | |
| - new_name | |
| responses: | |
| 200: | |
| description: return rename result | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| message: | |
| type: string | |
| /delete_folder: | |
| post: | |
| tags: | |
| - all | |
| summary: delete folder | |
| operationId: server.delete_folder | |
| parameters: | |
| - in: body | |
| name: delete_request | |
| description: request to delete folder | |
| schema: | |
| type: object | |
| properties: | |
| path: | |
| type: string | |
| description: folder path | |
| required: | |
| - path | |
| responses: | |
| 200: | |
| description: return delete result | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| message: | |
| type: string | |
| /list_all_folders: | |
| get: | |
| tags: | |
| - all | |
| summary: list all folders for move operation | |
| operationId: server.list_all_folders | |
| responses: | |
| 200: | |
| description: return list of all folders | |
| schema: | |
| type: object | |
| properties: | |
| folders: | |
| type: array | |
| items: | |
| type: string | |
| description: list of folder paths | |
| /create_folder: | |
| post: | |
| tags: | |
| - all | |
| summary: create a new folder | |
| operationId: server.create_folder_api | |
| parameters: | |
| - in: body | |
| name: create_request | |
| description: request to create folder | |
| schema: | |
| type: object | |
| properties: | |
| parent_path: | |
| type: string | |
| description: 'parent folder path (default: "/" for root directory)' | |
| default: "/" | |
| folder_name: | |
| type: string | |
| description: new folder name | |
| required: | |
| - folder_name | |
| responses: | |
| 200: | |
| description: return create result | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| message: | |
| type: string | |
| /fetch_url: | |
| post: | |
| tags: | |
| - all | |
| summary: fetch text content from URL | |
| operationId: server.fetch_url | |
| parameters: | |
| - in: body | |
| name: fetch_request | |
| description: request to fetch text from URL | |
| schema: | |
| type: object | |
| properties: | |
| url: | |
| type: string | |
| description: URL to fetch text from | |
| required: | |
| - url | |
| responses: | |
| 200: | |
| description: return fetched text content | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| text: | |
| type: string | |
| description: extracted text content | |
| url: | |
| type: string | |
| description: original URL | |
| char_count: | |
| type: integer | |
| description: character count of extracted text | |
| message: | |
| type: string | |
| description: error message if success is false | |
| 400: | |
| description: invalid request (missing url or invalid URL format) | |
| 500: | |
| description: failed to fetch or extract text from URL | |
| /client-activity: | |
| post: | |
| tags: | |
| - all | |
| summary: anonymous client activity heartbeat | |
| operationId: server.client_activity_report | |
| parameters: | |
| - in: body | |
| name: activity_body | |
| required: false | |
| schema: | |
| type: object | |
| properties: | |
| page_path: | |
| type: string | |
| description: pathname+search；汇总按去掉 ?/# 的路径末段，逐条日志里 path 仍带 query（unquote） | |
| total_active_sec: | |
| type: integer | |
| delta_active_sec: | |
| type: integer | |
| client_os: | |
| type: string | |
| description: 首轮心跳附带；ios / android / windows / macos / linux / unknown | |
| description: | | |
| 有效心跳计入访问统计与在线估计。请求头带有效 X-Admin-Token 时仅返回 online_now，不计入统计。 | |
| responses: | |
| 200: | |
| description: acknowledged; includes estimated online_now after valid heartbeat | |
| schema: | |
| type: object | |
| properties: | |
| ok: | |
| type: boolean | |
| online_now: | |
| type: integer | |
| online_window_sec: | |
| type: integer | |
| /check_admin: | |
| post: | |
| tags: | |
| - all | |
| summary: check admin token | |
| operationId: server.check_admin | |
| parameters: | |
| - in: body | |
| name: check_request | |
| description: request to check admin token | |
| schema: | |
| type: object | |
| properties: | |
| token: | |
| type: string | |
| description: admin token to check | |
| required: | |
| - token | |
| responses: | |
| 200: | |
| description: return check result | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| message: | |
| type: string | |
| /available_models: | |
| get: | |
| tags: | |
| - all | |
| summary: get available models | |
| operationId: server.get_available_models | |
| responses: | |
| 200: | |
| description: return list of available models | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| models: | |
| type: array | |
| items: | |
| type: string | |
| /current_model: | |
| get: | |
| tags: | |
| - all | |
| summary: get current model | |
| operationId: server.get_current_model | |
| responses: | |
| 200: | |
| description: return current model name and quantization config | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| model: | |
| type: string | |
| switching: | |
| type: boolean | |
| device_type: | |
| type: string | |
| enum: [cpu, cuda, mps] | |
| description: current device type | |
| use_int8: | |
| type: boolean | |
| description: whether INT8 quantization is enabled | |
| use_bfloat16: | |
| type: boolean | |
| description: whether bfloat16 is enabled (CPU only) | |
| /prediction-attribute: | |
| post: | |
| tags: | |
| - all | |
| summary: prediction attribute | |
| operationId: server.prediction_attribute | |
| parameters: | |
| - in: body | |
| name: attribution_request | |
| schema: | |
| type: object | |
| properties: | |
| context: | |
| type: string | |
| description: 输入上下文文本（不做 trim；token 数不得超过 2000，否则 400） | |
| target_prediction: | |
| type: string | |
| description: 目标预测文本，取首个 token 作为归因目标（不做 trim）。省略时自动使用 top-1（贪心解码）。 | |
| model: | |
| type: string | |
| enum: [base, instruct] | |
| description: base 使用主槽位（--base_model），instruct 使用 instruct 槽位（--instruct_model） | |
| source_page: | |
| type: string | |
| description: 发起页面 slug（analysis / chat / attribution / causal_flow） | |
| flow_id: | |
| type: string | |
| description: 连续生成归因会话 ID；仅 source_page=causal_flow 时允许 | |
| flow_step: | |
| type: integer | |
| description: 连续生成归因步骤（从 0 开始）；仅 source_page=causal_flow 时允许 | |
| required: | |
| - context | |
| - model | |
| - source_page | |
| responses: | |
| 200: | |
| description: 返回各输入 token 对目标预测的归因分 | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| model: | |
| type: string | |
| target_token: | |
| type: string | |
| description: 归因目标 token 的字符串 | |
| target_prob: | |
| type: number | |
| description: 该 token 在 next-token 分布中的预测概率 | |
| token_attribution: | |
| type: array | |
| items: | |
| type: object | |
| properties: | |
| offset: | |
| type: array | |
| items: | |
| type: integer | |
| description: 字符偏移 [start, end] | |
| raw: | |
| type: string | |
| description: token 原文 | |
| score: | |
| type: number | |
| description: 梯度 L2 范数归因分 | |
| debug_info: | |
| type: object | |
| description: 下一 token 的 top10 与概率（与语义分析 debug_info 同形，无 abbrev） | |
| properties: | |
| topk_tokens: | |
| type: array | |
| items: | |
| type: string | |
| topk_probs: | |
| type: array | |
| items: | |
| type: number | |
| is_eos: | |
| type: boolean | |
| description: target_token 是否为模型的 EOS token（top-1 模式下尤为重要，客户端据此判断是否终止生成循环） | |
| 400: | |
| description: 缺少必要字段、model 非法、context 超长、或 target 无法 tokenize | |
| 500: | |
| description: 推理失败 | |
| 503: | |
| description: 服务繁忙 | |
| /ablation-attribute: | |
| post: | |
| tags: | |
| - all | |
| summary: ablation attribution | |
| operationId: server.ablation_attribute | |
| parameters: | |
| - in: body | |
| name: ablation_request | |
| schema: | |
| type: object | |
| required: | |
| - context | |
| - model | |
| - source_page | |
| properties: | |
| context: | |
| type: string | |
| description: 输入上下文文本（token 数不得超过 500，否则 400） | |
| target_prediction: | |
| type: string | |
| description: 归因目标文本，取首个 token。省略时自动使用 top-1（贪心解码）。 | |
| target_token_id: | |
| type: integer | |
| description: 归因目标 token id（与 target_prediction 互斥）。 | |
| model: | |
| type: string | |
| enum: [base, instruct] | |
| description: base 使用主槽位，instruct 使用 instruct 槽位 | |
| source_page: | |
| type: string | |
| description: 发起页面 slug（analysis / chat / attribution / causal_flow） | |
| flow_id: | |
| type: string | |
| description: 连续归因会话 ID；仅 source_page=causal_flow 时允许 | |
| flow_step: | |
| type: integer | |
| description: 连续归因步骤（从 0 开始）；仅 source_page=causal_flow 时允许 | |
| responses: | |
| 200: | |
| description: 返回各输入 token 对目标预测的消融归因分 | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| model: | |
| type: string | |
| target_token: | |
| type: string | |
| description: 归因目标 token 的字符串 | |
| target_prob: | |
| type: number | |
| description: baseline 下该 token 的预测概率 | |
| token_attribution: | |
| type: array | |
| items: | |
| type: object | |
| properties: | |
| offset: | |
| type: array | |
| items: | |
| type: integer | |
| description: 字符偏移 [start, end] | |
| raw: | |
| type: string | |
| description: token 原文 | |
| score: | |
| type: number | |
| description: ΔP = baseline_prob − occluded_prob（可为负） | |
| delta_logit: | |
| type: number | |
| description: Δlogit = baseline_logit − occluded_logit（可选补充） | |
| debug_info: | |
| type: object | |
| properties: | |
| topk_tokens: | |
| type: array | |
| items: | |
| type: string | |
| topk_probs: | |
| type: array | |
| items: | |
| type: number | |
| is_eos: | |
| type: boolean | |
| description: target_token 是否为 EOS token | |
| 400: | |
| description: 缺少必要字段、model 非法、context 超长、target 互斥或无法 tokenize | |
| 500: | |
| description: 推理失败 | |
| 503: | |
| description: 服务繁忙 | |
| /logit-lens: | |
| post: | |
| tags: | |
| - all | |
| summary: logit lens | |
| operationId: server.logit_lens | |
| parameters: | |
| - in: body | |
| name: logit_lens_request | |
| schema: | |
| type: object | |
| required: | |
| - context | |
| - model | |
| - source_page | |
| properties: | |
| context: | |
| type: string | |
| description: 输入上下文文本（token 数不得超过 500，否则 400） | |
| target_prediction: | |
| type: string | |
| description: 归因目标文本，取首个 token。省略时自动使用 top-1。 | |
| target_token_id: | |
| type: integer | |
| description: 归因目标 token id（与 target_prediction 互斥）。 | |
| model: | |
| type: string | |
| enum: [base, instruct] | |
| source_page: | |
| type: string | |
| flow_id: | |
| type: string | |
| flow_step: | |
| type: integer | |
| responses: | |
| 200: | |
| description: 逐层 Logit Lens 投影结果 | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| model: | |
| type: string | |
| target_token: | |
| type: string | |
| n_layers: | |
| type: integer | |
| description: Transformer 层数 L（不含 embedding 层） | |
| final_target_prob: | |
| type: number | |
| description: 最终层目标 token 概率 | |
| layers: | |
| type: array | |
| items: | |
| type: object | |
| properties: | |
| layer: | |
| type: integer | |
| is_embedding: | |
| type: boolean | |
| topk_tokens: | |
| type: array | |
| items: | |
| type: string | |
| topk_probs: | |
| type: array | |
| items: | |
| type: number | |
| target_prob: | |
| type: number | |
| debug_info: | |
| type: object | |
| properties: | |
| topk_tokens: | |
| type: array | |
| items: | |
| type: string | |
| topk_probs: | |
| type: array | |
| items: | |
| type: number | |
| is_eos: | |
| type: boolean | |
| 400: | |
| description: 缺少必要字段、model 非法、context 超长、target 互斥 | |
| 500: | |
| description: 推理失败 | |
| 503: | |
| description: 服务繁忙 | |
| /activation-explain: | |
| post: | |
| tags: | |
| - all | |
| summary: activation explain (Tiny-NLA) | |
| operationId: server.activation_explain | |
| parameters: | |
| - in: body | |
| name: activation_explain_request | |
| schema: | |
| type: object | |
| required: | |
| - model | |
| - source_page | |
| properties: | |
| model: | |
| type: string | |
| enum: [base, instruct] | |
| source_page: | |
| type: string | |
| text: | |
| type: string | |
| description: 输入上下文文本（与 token_index 配套） | |
| token_index: | |
| type: integer | |
| description: 目标 token 索引（text 模式必填） | |
| vector: | |
| type: array | |
| items: | |
| type: number | |
| description: 1024 维激活向量（与 text 互斥） | |
| responses: | |
| 200: | |
| description: 激活解释结果 | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| concept: | |
| type: string | |
| description: 激活概念标签 | |
| explanation: | |
| type: string | |
| description: 自然语言解释 | |
| roundtrip_cosine: | |
| type: number | |
| description: roundtrip 可信度（0-1） | |
| vector_dim: | |
| type: integer | |
| description: 向量维度 | |
| note: | |
| type: string | |
| description: 备注或状态说明 | |
| 400: | |
| description: 缺少必要字段、model 非法、text/vector 均未提供 | |
| 500: | |
| description: 推理失败 | |
| 503: | |
| description: 服务繁忙 | |
| /branch-next: | |
| post: | |
| tags: | |
| - all | |
| summary: branch next token candidates | |
| operationId: server.branch_next | |
| parameters: | |
| - in: body | |
| name: branch_next_request | |
| schema: | |
| type: object | |
| required: | |
| - prefix | |
| - model | |
| - source_page | |
| properties: | |
| prefix: | |
| type: string | |
| description: 当前路径的前缀文本 | |
| model: | |
| type: string | |
| enum: [base, instruct] | |
| source_page: | |
| type: string | |
| top_k: | |
| type: integer | |
| description: 返回候选数（默认 10，上限 50，超出自动 clamp） | |
| responses: | |
| 200: | |
| description: 下一 token 的 top-k 候选 | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| model: | |
| type: string | |
| prefix_tokens: | |
| type: integer | |
| description: prefix 的 token 数 | |
| candidates: | |
| type: array | |
| items: | |
| type: object | |
| properties: | |
| token: | |
| type: string | |
| token_id: | |
| type: integer | |
| prob: | |
| type: number | |
| is_context_full: | |
| type: boolean | |
| description: prefix 是否已占满上下文（前端据此禁用展开） | |
| 400: | |
| description: 缺少必要字段、model 非法、prefix 超出上下文上限 | |
| 500: | |
| description: 推理失败 | |
| 503: | |
| description: 服务繁忙 | |
| /tokenize: | |
| post: | |
| tags: | |
| - all | |
| summary: tokenize text | |
| description: | | |
| 对 context 用指定 model 的 tokenizer 分词，返回各 token 的字符 offset 与原文。 | |
| 不持有推理锁，不做前向 / 梯度计算，响应极快。 | |
| operationId: server.tokenize | |
| parameters: | |
| - in: body | |
| name: tokenize_request | |
| required: true | |
| schema: | |
| type: object | |
| required: | |
| - context | |
| - model | |
| properties: | |
| context: | |
| type: string | |
| description: 待分词文本 | |
| model: | |
| type: string | |
| enum: [base, instruct] | |
| description: base 使用主槽位 tokenizer，instruct 使用语义槽位 tokenizer | |
| responses: | |
| 200: | |
| description: 分词结果 | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| spans: | |
| type: array | |
| items: | |
| type: object | |
| properties: | |
| offset: | |
| type: array | |
| items: | |
| type: integer | |
| description: 字符偏移 [start, end] | |
| raw: | |
| type: string | |
| description: token 原文 | |
| 400: | |
| description: 缺少必要字段或 model 非法 | |
| /analyze-semantic: | |
| post: | |
| tags: | |
| - all | |
| summary: analyze semantic (token-level relevance to query) | |
| operationId: server.analyze_semantic | |
| parameters: | |
| - in: body | |
| name: semantic_request | |
| schema: | |
| type: object | |
| properties: | |
| query: | |
| type: string | |
| description: 查询主题 | |
| text: | |
| type: string | |
| description: 原文 | |
| stream: | |
| type: boolean | |
| description: 是否使用 SSE 流式响应（带阶段级进度） | |
| submode: | |
| type: string | |
| enum: [count, match_score, fill_blank] | |
| description: 可选子模式，不传则用服务端默认。match_score 已废弃 | |
| debug_info: | |
| type: boolean | |
| description: 为 true 时响应包含 debug_info（abbrev、topk_tokens、topk_probs），默认关，不传则不返回 | |
| required: | |
| - query | |
| - text | |
| responses: | |
| 200: | |
| description: 返回 model 和 token_attention | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| model: | |
| type: string | |
| description: 使用的 semantic 模型名 | |
| token_attention: | |
| type: array | |
| items: | |
| type: object | |
| properties: | |
| offset: | |
| type: array | |
| items: | |
| type: integer | |
| description: 原文中的字符偏移 [start, end] | |
| raw: | |
| type: string | |
| description: token 原文 | |
| score: | |
| type: number | |
| description: 对 prompt 区域的平均关注度 | |
| debug_info: | |
| type: object | |
| description: 请求中 debug_info=true 时返回 | |
| properties: | |
| abbrev: | |
| type: string | |
| description: 推理原文缩写 | |
| topk_tokens: | |
| type: array | |
| items: | |
| type: string | |
| description: top10 预测 token 列表 | |
| topk_probs: | |
| type: array | |
| items: | |
| type: number | |
| description: top10 对应的概率 | |
| message: | |
| type: string | |
| 400: | |
| description: 缺少 query 或 text 参数 | |
| 500: | |
| description: 分析失败 | |
| /switch_model: | |
| post: | |
| tags: | |
| - all | |
| summary: switch model (admin only) | |
| operationId: server.switch_model | |
| parameters: | |
| - in: body | |
| name: switch_request | |
| description: request to switch model with quantization options | |
| schema: | |
| type: object | |
| properties: | |
| model: | |
| type: string | |
| description: target model name | |
| use_int8: | |
| type: boolean | |
| description: enable INT8 quantization (not supported on MPS) | |
| default: false | |
| use_bfloat16: | |
| type: boolean | |
| description: enable bfloat16 (CPU only) | |
| default: false | |
| required: | |
| - model | |
| responses: | |
| 200: | |
| description: return switch result | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| message: | |
| type: string | |
| model: | |
| type: string | |
| 400: | |
| description: invalid request or incompatible quantization settings | |
| 404: | |
| description: model not found | |
| 500: | |
| description: model switch failed | |
| 503: | |
| description: model switching in progress | |
| /v1/completions/prompt: | |
| post: | |
| tags: | |
| - openai | |
| summary: Apply chat template to user text (completion prompt assembly) | |
| description: | | |
| 将 messages 经 tokenizer.apply_chat_template 封装，返回实际送入 POST /v1/completions 的完整字符串（JSON 字段 prompt_used）。 | |
| 可选 tools、enable_thinking。与续写共用 semantic 模型/tokenizer；过长时返回 400。 | |
| operationId: server.completions_prompt | |
| parameters: | |
| - in: body | |
| name: completions_prompt_request | |
| required: true | |
| schema: | |
| $ref: "server_openai_definitions.yaml#/definitions/OpenAICompletionsPromptRequest" | |
| responses: | |
| 200: | |
| description: ok | |
| schema: | |
| $ref: "server_openai_definitions.yaml#/definitions/OpenAICompletionsPromptResponse" | |
| 400: | |
| description: invalid body or prompt too long | |
| /v1/completions/prompt-incremental: | |
| post: | |
| tags: | |
| - openai | |
| summary: Compute wire incremental suffix for a tool response | |
| description: | | |
| 多轮 wire 模式下，计算追加 tool response 及下一轮 generation scaffold 的字面量后缀（incremental_suffix）。 | |
| 仅依赖 tool_content 和 enable_thinking，与前序 wire 历史内容无关。 | |
| operationId: server.completions_prompt_incremental | |
| parameters: | |
| - in: body | |
| name: completions_prompt_incremental_request | |
| required: true | |
| schema: | |
| $ref: "server_openai_definitions.yaml#/definitions/OpenAICompletionsPromptIncrementalRequest" | |
| responses: | |
| 200: | |
| description: ok | |
| schema: | |
| $ref: "server_openai_definitions.yaml#/definitions/OpenAICompletionsPromptIncrementalResponse" | |
| 400: | |
| description: invalid body | |
| 500: | |
| description: template extraction failed | |
| /v1/completions: | |
| post: | |
| tags: | |
| - openai | |
| summary: OpenAI-compatible text completion (minimal v1) | |
| description: | | |
| 与 OpenAI POST /v1/completions 对齐（见 platform.openai.com Completions）。 | |
| 必填 model、prompt；prompt 须为已确定的完整模型输入（需 chat template 时请先 POST /v1/completions/prompt）。 | |
| 本接口响应恒为 text/event-stream（忽略 body 中的 stream 字段）：type=delta…，末条 type=result，其 data 与 OpenAICompletionsResponse 同形（含 choices[0].text、info_radar.bpe_strings）；若需「非流式」语义，由客户端消费 SSE 末条 result 即可。 | |
| operationId: server.completions | |
| parameters: | |
| - in: body | |
| name: completions_request | |
| required: true | |
| schema: | |
| $ref: "server_openai_definitions.yaml#/definitions/OpenAICompletionsRequest" | |
| responses: | |
| 200: | |
| description: | | |
| text/event-stream（SSE）；末条事件的 data 与 OpenAICompletionsResponse 同形。 | |
| schema: | |
| $ref: "server_openai_definitions.yaml#/definitions/OpenAICompletionsResponse" | |
| 400: | |
| description: invalid request body | |
| /v1/completions/stop: | |
| post: | |
| tags: | |
| - openai | |
| summary: Global completions stop (single-user serial) | |
| description: | | |
| 置位全局停止标志，使当前正在进行的续写尽快结束（与单次 SSE 断开/GeneratorExit 叠加）。 | |
| 单用户串行场景下无需 request_id；新一次 POST /v1/completions 开始时清除该标志。 | |
| operationId: server.completions_stop | |
| responses: | |
| 200: | |
| description: ok | |
| schema: | |
| type: object | |
| properties: | |
| ok: | |
| type: boolean | |
| /visit_stats: | |
| get: | |
| tags: | |
| - all | |
| summary: get visit statistics (admin only) | |
| operationId: server.get_visit_stats | |
| responses: | |
| 200: | |
| description: | | |
| 返回当前累计快照（_base + 本进程内存增量）。 | |
| 窗口内未提交的增量仅以与 stats_total.json 相同字段结构的记录写入 Dataset 的 stats_delta/*.json。 | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| totals: | |
| type: object | |
| description: | | |
| page_loads（GET /client/*.html，一次实际 HTML 响应计一次，与访问日志「📄 页面访问」一致）；active_visits（至少有一次有效活跃心跳的页面访问数，每页以 delta_active_sec == total_active_sec 的首轮上报计一次）。 | |
| os: | |
| type: object | |
| page_sec: | |
| type: object | |
| api: | |
| type: object | |
| saved_at: | |
| type: string | |
| startup_base: | |
| type: object | |
| description: | | |
| 进程启动时（_load_base 完成后）_base 的深拷贝；前端用 totals − startup_base 计算自启动以来的增量。 | |
| process_start_at: | |
| type: string | |
| description: | | |
| 统计线程完成 _load_base 并拍下 startup_base 时的 UTC 时间（与 saved_at 相同格式）。 | |
| reset_base: | |
| type: object | |
| description: | | |
| 最近一次手动 reset 时的全量快照；前端用 totals − reset_base 计算自 reset 以来的增量。 | |
| reset_at: | |
| type: string | |
| description: | | |
| 最近一次手动 reset 的 UTC 时间；从未 reset 时为 null。 | |
| online_now: | |
| type: integer | |
| description: | | |
| 最近 online_window_sec 内 client-activity 心跳数按每 10s 一次折算的当前在线估计（ceil）；进程内存，重启清零。 | |
| online_window_sec: | |
| type: integer | |
| description: 在线估计滑动窗口秒数（当前为 20）。 | |
| /visit_stats/active_visits_timeline: | |
| get: | |
| tags: | |
| - all | |
| summary: active visits and active time per UTC hour (admin only) | |
| operationId: server.get_visit_stats_active_visits_timeline | |
| responses: | |
| 200: | |
| description: hourly bins from stats_delta and unpersisted session increment | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| bins: | |
| type: array | |
| items: | |
| type: object | |
| properties: | |
| hour: | |
| type: string | |
| description: UTC hour start, e.g. 2025-06-03T12:00:00Z | |
| active_visits: | |
| type: integer | |
| active_sec: | |
| type: integer | |
| description: sum of page active seconds in this UTC hour bucket | |
| 503: | |
| description: HF token missing or Hub unavailable | |
| /visit_stats/reset: | |
| post: | |
| tags: | |
| - all | |
| summary: reset delta base (admin only) | |
| operationId: server.post_visit_stats_reset | |
| responses: | |
| 200: | |
| description: reset base saved | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| 500: | |
| description: persist failed | |
| schema: | |
| type: object | |
| properties: | |
| success: | |
| type: boolean | |
| error: | |
| type: string | |
| definitions: | |
| # 与 client generatedSchemas TokenWithOffset 一致；信息密度 analyze 与 completions info_radar 共用 | |
| TokenWithOffset: | |
| type: object | |
| properties: | |
| offset: | |
| type: array | |
| description: | | |
| 字符下标 [start, end)：信息密度分析相对 request.text； | |
| completions 的 info_radar 相对续写全文 choices[0].text。 | |
| items: | |
| type: integer | |
| minItems: 2 | |
| maxItems: 2 | |
| raw: | |
| type: string | |
| real_topk: | |
| type: array | |
| description: "[模型排序名次, 该 token 的 softmax 概率]" | |
| items: | |
| type: number | |
| minItems: 2 | |
| maxItems: 2 | |
| pred_topk: | |
| type: array | |
| description: 该位置 logits softmax 后的 top-N 候选 [token 文本, 概率]（N 由服务端 DEFAULT_TOPK 决定） | |
| # 与 client generatedSchemas AnalyzeResult 一致；成功路径下 backend.api.analyze._build_response | |
| AnalyzeResult: | |
| type: object | |
| required: | |
| - bpe_strings | |
| properties: | |
| model: | |
| type: string | |
| bpe_strings: | |
| type: array | |
| items: | |
| $ref: "#/definitions/TokenWithOffset" | |
| error: | |
| type: string | |
| x-nullable: true | |
| AnalyzeResponse: | |
| type: object | |
| required: | |
| - request | |
| - result | |
| properties: | |
| request: | |
| type: object | |
| required: | |
| - text | |
| properties: | |
| text: | |
| type: string | |
| description: 与请求中分析的原文一致（响应中仅回显 text） | |
| result: | |
| $ref: "#/definitions/AnalyzeResult" | |
| AnalyzeRequest: | |
| type: object | |
| required: | |
| - model | |
| - text | |
| properties: | |
| model: | |
| type: string | |
| text: | |
| type: string | |
| stream: | |
| type: boolean | |
| description: 是否使用 SSE 流式响应 | |