docs: comprehensive docs review + i18n sync

- Fix .gitignore: add a2a-server.md, auto-combo.md, mcp-server.md, new-features/ to whitelist
- Rewrite FEATURES.md: 18 sections covering v2.0.12 state (Playground, Themes, CLI Agents, Media, API Keys, Audit Log)
- API_REFERENCE.md: add ACP Agents endpoints (/api/acp/agents GET/POST/DELETE)
- Sync all 6 root docs to 29 i18n directories (174 files)
- Remove stale git-tracked docs (adr/, i18n-tasks/)
This commit is contained in:
diegosouzapw 2026-03-07 12:18:17 -03:00
parent 2306081dab
commit 91f3bd4056
210 changed files with 37754 additions and 31332 deletions

View file

@ -1,22 +1,22 @@
# omniroute — Dokumentasi Pangkalan Kod
# omniroute — Codebase Documentation
🌐 **Languages:** 🇺🇸 [English](../../CODEBASE_DOCUMENTATION.md) | 🇧🇷 [Português (Brasil)](../pt-BR/CODEBASE_DOCUMENTATION.md) | 🇪🇸 [Español](../es/CODEBASE_DOCUMENTATION.md) | 🇫🇷 [Français](../fr/CODEBASE_DOCUMENTATION.md) | 🇮🇹 [Italiano](../it/CODEBASE_DOCUMENTATION.md) | 🇷🇺 [Русский](../ru/CODEBASE_DOCUMENTATION.md) | 🇨🇳 [中文 (简体)](../zh-CN/CODEBASE_DOCUMENTATION.md) | 🇩🇪 [Deutsch](../de/CODEBASE_DOCUMENTATION.md) | 🇮🇳 [हिन्दी](../in/CODEBASE_DOCUMENTATION.md) | 🇹🇭 [ไทย](../th/CODEBASE_DOCUMENTATION.md) | 🇺🇦 [Українська](../uk-UA/CODEBASE_DOCUMENTATION.md) | 🇸🇦 [العربية](../ar/CODEBASE_DOCUMENTATION.md) | 🇯🇵 [日本語](../ja/CODEBASE_DOCUMENTATION.md) | 🇻🇳 [Tiếng Việt](../vi/CODEBASE_DOCUMENTATION.md) | 🇧🇬 [Български](../bg/CODEBASE_DOCUMENTATION.md) | 🇩🇰 [Dansk](../da/CODEBASE_DOCUMENTATION.md) | 🇫🇮 [Suomi](../fi/CODEBASE_DOCUMENTATION.md) | 🇮🇱 [עברית](../he/CODEBASE_DOCUMENTATION.md) | 🇭🇺 [Magyar](../hu/CODEBASE_DOCUMENTATION.md) | 🇮🇩 [Bahasa Indonesia](../id/CODEBASE_DOCUMENTATION.md) | 🇰🇷 [한국어](../ko/CODEBASE_DOCUMENTATION.md) | 🇲🇾 [Bahasa Melayu](../ms/CODEBASE_DOCUMENTATION.md) | 🇳🇱 [Nederlands](../nl/CODEBASE_DOCUMENTATION.md) | 🇳🇴 [Norsk](../no/CODEBASE_DOCUMENTATION.md) | 🇵🇹 [Português (Portugal)](../pt/CODEBASE_DOCUMENTATION.md) | 🇷🇴 [Română](../ro/CODEBASE_DOCUMENTATION.md) | 🇵🇱 [Polski](../pl/CODEBASE_DOCUMENTATION.md) | 🇸🇰 [Slovenčina](../sk/CODEBASE_DOCUMENTATION.md) | 🇸🇪 [Svenska](../sv/CODEBASE_DOCUMENTATION.md) | 🇵🇭 [Filipino](../phi/CODEBASE_DOCUMENTATION.md)
🌐 **Languages:** 🇺🇸 [English](CODEBASE_DOCUMENTATION.md) | 🇧🇷 [Português (Brasil)](i18n/pt-BR/CODEBASE_DOCUMENTATION.md) | 🇪🇸 [Español](i18n/es/CODEBASE_DOCUMENTATION.md) | 🇫🇷 [Français](i18n/fr/CODEBASE_DOCUMENTATION.md) | 🇮🇹 [Italiano](i18n/it/CODEBASE_DOCUMENTATION.md) | 🇷🇺 [Русский](i18n/ru/CODEBASE_DOCUMENTATION.md) | 🇨🇳 [中文 (简体)](i18n/zh-CN/CODEBASE_DOCUMENTATION.md) | 🇩🇪 [Deutsch](i18n/de/CODEBASE_DOCUMENTATION.md) | 🇮🇳 [हिन्दी](i18n/in/CODEBASE_DOCUMENTATION.md) | 🇹🇭 [ไทย](i18n/th/CODEBASE_DOCUMENTATION.md) | 🇺🇦 [Українська](i18n/uk-UA/CODEBASE_DOCUMENTATION.md) | 🇸🇦 [العربية](i18n/ar/CODEBASE_DOCUMENTATION.md) | 🇯🇵 [日本語](i18n/ja/CODEBASE_DOCUMENTATION.md) | 🇻🇳 [Tiếng Việt](i18n/vi/CODEBASE_DOCUMENTATION.md) | 🇧🇬 [Български](i18n/bg/CODEBASE_DOCUMENTATION.md) | 🇩🇰 [Dansk](i18n/da/CODEBASE_DOCUMENTATION.md) | 🇫🇮 [Suomi](i18n/fi/CODEBASE_DOCUMENTATION.md) | 🇮🇱 [עברית](i18n/he/CODEBASE_DOCUMENTATION.md) | 🇭🇺 [Magyar](i18n/hu/CODEBASE_DOCUMENTATION.md) | 🇮🇩 [Bahasa Indonesia](i18n/id/CODEBASE_DOCUMENTATION.md) | 🇰🇷 [한국어](i18n/ko/CODEBASE_DOCUMENTATION.md) | 🇲🇾 [Bahasa Melayu](i18n/ms/CODEBASE_DOCUMENTATION.md) | 🇳🇱 [Nederlands](i18n/nl/CODEBASE_DOCUMENTATION.md) | 🇳🇴 [Norsk](i18n/no/CODEBASE_DOCUMENTATION.md) | 🇵🇹 [Português (Portugal)](i18n/pt/CODEBASE_DOCUMENTATION.md) | 🇷🇴 [Română](i18n/ro/CODEBASE_DOCUMENTATION.md) | 🇵🇱 [Polski](i18n/pl/CODEBASE_DOCUMENTATION.md) | 🇸🇰 [Slovenčina](i18n/sk/CODEBASE_DOCUMENTATION.md) | 🇸🇪 [Svenska](i18n/sv/CODEBASE_DOCUMENTATION.md) | 🇵🇭 [Filipino](i18n/phi/CODEBASE_DOCUMENTATION.md)
> Panduan komprehensif dan mesra pemula kepada penghala proksi AI **omniroute** berbilang pembekal.
> A comprehensive, beginner-friendly guide to the **omniroute** multi-provider AI proxy router.
---
## 1. Apakah itu omniroute?
## 1. What Is omniroute?
omniroute ialah **penghala proksi** yang terletak di antara klien AI (Claude CLI, Codex, Cursor IDE, dll.) dan penyedia AI (Anthropic, Google, OpenAI, AWS, GitHub, dsb.). Ia menyelesaikan satu masalah besar:
omniroute is a **proxy router** that sits between AI clients (Claude CLI, Codex, Cursor IDE, etc.) and AI providers (Anthropic, Google, OpenAI, AWS, GitHub, etc.). It solves one big problem:
> **Pelanggan AI yang berbeza bercakap "bahasa" yang berbeza (format API), dan pembekal AI yang berbeza juga mengharapkan "bahasa" yang berbeza.** omniroute menterjemah antara mereka secara automatik.
> **Different AI clients speak different "languages" (API formats), and different AI providers expect different "languages" too.** omniroute translates between them automatically.
Anggaplah ia seperti penterjemah universal di Pertubuhan Bangsa-Bangsa Bersatu — mana-mana perwakilan boleh bercakap apa-apa bahasa, dan penterjemah menukarnya untuk mana-mana perwakilan lain.
Think of it like a universal translator at the United Nations — any delegate can speak any language, and the translator converts it for any other delegate.
---
## 2. Gambaran Keseluruhan Seni Bina
## 2. Architecture Overview
```mermaid
graph LR
@ -61,20 +61,20 @@ graph LR
H -.-> G
```
### Prinsip Teras: Terjemahan Hub-and-Spoke
### Core Principle: Hub-and-Spoke Translation
Semua terjemahan format melalui **format OpenAI sebagai hab**:
All format translation passes through **OpenAI format as the hub**:
```
Client Format → [OpenAI Hub] → Provider Format (request)
Provider Format → [OpenAI Hub] → Client Format (response)
```
Ini bermakna anda hanya memerlukan **N penterjemah** (satu setiap format) dan bukannya **N²** (setiap pasangan).
This means you only need **N translators** (one per format) instead of **N²** (every pair).
---
## 3. Struktur Projek
## 3. Project Structure
```
omniroute/
@ -104,22 +104,22 @@ omniroute/
---
## 4. Pecahan Modul demi Modul
## 4. Module-by-Module Breakdown
### 4.1 Konfigurasi (`open-sse/config/`)
### 4.1 Config (`open-sse/config/`)
**Sumber tunggal kebenaran** untuk semua konfigurasi pembekal.
The **single source of truth** for all provider configuration.
| Fail | Tujuan |
| ----------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `constants.ts` | `PROVIDERS` objek dengan URL asas, bukti kelayakan OAuth (lalai), pengepala dan gesaan sistem lalai untuk setiap pembekal. Juga mentakrifkan `HTTP_STATUS`, `ERROR_TYPES`, `COOLDOWN_MS`, `BACKOFF_CONFIG` dan `SKIP_PATTERNS`. |
| `credentialLoader.ts` | Memuatkan bukti kelayakan luaran daripada `data/provider-credentials.json` dan menggabungkannya pada lalai berkod keras dalam `PROVIDERS`. Menyimpan rahsia di luar kawalan sumber sambil mengekalkan keserasian ke belakang. |
| `providerModels.ts` | Pendaftaran model pusat: alias penyedia peta → ID model. Berfungsi seperti `getModels()`, `getProviderByAlias()`. |
| `codexInstructions.ts` | Arahan sistem disuntik ke dalam permintaan Codex (kekangan pengeditan, peraturan kotak pasir, dasar kelulusan). |
| `defaultThinkingSignature.ts` | Tanda tangan "berfikir" lalai untuk model Claude dan Gemini. |
| `ollamaModels.ts` | Takrif skema untuk model Ollama tempatan (nama, saiz, keluarga, pengkuantitian). |
| File | Purpose |
| ----------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `constants.ts` | `PROVIDERS` object with base URLs, OAuth credentials (defaults), headers, and default system prompts for every provider. Also defines `HTTP_STATUS`, `ERROR_TYPES`, `COOLDOWN_MS`, `BACKOFF_CONFIG`, and `SKIP_PATTERNS`. |
| `credentialLoader.ts` | Loads external credentials from `data/provider-credentials.json` and merges them over the hardcoded defaults in `PROVIDERS`. Keeps secrets out of source control while maintaining backwards compatibility. |
| `providerModels.ts` | Central model registry: maps provider aliases → model IDs. Functions like `getModels()`, `getProviderByAlias()`. |
| `codexInstructions.ts` | System instructions injected into Codex requests (editing constraints, sandbox rules, approval policies). |
| `defaultThinkingSignature.ts` | Default "thinking" signatures for Claude and Gemini models. |
| `ollamaModels.ts` | Schema definition for local Ollama models (name, size, family, quantization). |
#### Aliran Pemuatan Kredensial
#### Credential Loading Flow
```mermaid
flowchart TD
@ -142,9 +142,9 @@ flowchart TD
---
### 4.2 Pelaksana (`open-sse/executors/`)
### 4.2 Executors (`open-sse/executors/`)
Pelaksana merangkum **logik khusus pembekal** menggunakan **Corak Strategi**. Setiap pelaksana mengatasi kaedah asas seperti yang diperlukan.
Executors encapsulate **provider-specific logic** using the **Strategy Pattern**. Each executor overrides base methods as needed.
```mermaid
classDiagram
@ -194,32 +194,32 @@ classDiagram
BaseExecutor <|-- GithubExecutor
```
| Pelaksana | Pembekal | Pengkhususan Utama |
| ---------------- | ------------------------------------------ | -------------------------------------------------------------------------------------------------------------------------------------- |
| `base.ts` | — | Asas abstrak: Pembinaan URL, pengepala, cuba semula logik, penyegaran semula kelayakan |
| `default.ts` | Claude, Gemini, OpenAI, GLM, Kimi, MiniMax | Muat semula token OAuth generik untuk pembekal standard |
| `antigravity.ts` | Kod Awan Google | Penjanaan ID projek/sesi, sandaran berbilang URL, cuba semula tersuai menghuraikan daripada mesej ralat ("set semula selepas 2h7m23s") |
| `cursor.ts` | IDE kursor | **Paling kompleks**: Pengesahan checksum SHA-256, pengekodan permintaan Protobuf, Perduaan EventStream → Penghuraian respons SSE |
| `codex.ts` | OpenAI Codex | Menyuntik arahan sistem, mengurus tahap pemikiran, mengalih keluar parameter yang tidak disokong |
| `gemini-cli.ts` | Google Gemini CLI | Pembinaan URL tersuai (`streamGenerateContent`), muat semula token Google OAuth |
| `github.ts` | GitHub Copilot | Sistem token dwi (GitHub OAuth + Copilot token), pengepala VSCode meniru |
| `kiro.ts` | AWS CodeWhisperer | Penghuraian binari AWS EventStream, bingkai acara AMZN, anggaran token |
| `index.ts` | — | Kilang: nama pembekal peta → kelas pelaksana, dengan sandaran lalai |
| Executor | Provider | Key Specializations |
| ---------------- | ------------------------------------------ | ------------------------------------------------------------------------------------------------------------------- |
| `base.ts` | — | Abstract base: URL building, headers, retry logic, credential refresh |
| `default.ts` | Claude, Gemini, OpenAI, GLM, Kimi, MiniMax | Generic OAuth token refresh for standard providers |
| `antigravity.ts` | Google Cloud Code | Project/session ID generation, multi-URL fallback, custom retry parsing from error messages ("reset after 2h7m23s") |
| `cursor.ts` | Cursor IDE | **Most complex**: SHA-256 checksum auth, Protobuf request encoding, binary EventStream → SSE response parsing |
| `codex.ts` | OpenAI Codex | Injects system instructions, manages thinking levels, removes unsupported parameters |
| `gemini-cli.ts` | Google Gemini CLI | Custom URL building (`streamGenerateContent`), Google OAuth token refresh |
| `github.ts` | GitHub Copilot | Dual token system (GitHub OAuth + Copilot token), VSCode header mimicking |
| `kiro.ts` | AWS CodeWhisperer | AWS EventStream binary parsing, AMZN event frames, token estimation |
| `index.ts` | — | Factory: maps provider name → executor class, with default fallback |
---
### 4.3 Pengendali (`open-sse/handlers/`)
### 4.3 Handlers (`open-sse/handlers/`)
**Lapisan orkestrasi** — menyelaras terjemahan, pelaksanaan, penstriman dan pengendalian ralat.
The **orchestration layer** — coordinates translation, execution, streaming, and error handling.
| Fail | Tujuan |
| --------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `chatCore.ts` | **Orkestra pusat** (~600 baris). Mengendalikan kitaran hayat permintaan yang lengkap: pengesanan format → terjemahan → penghantaran pelaksana → respons penstriman/bukan penstriman → penyegaran token → pengendalian ralat → pengelogan penggunaan. |
| `responsesHandler.ts` | Penyesuai untuk API Respons OpenAI: menukar format Respons → Selesai Sembang → menghantar kepada `chatCore` → menukar SSE kembali kepada format Respons. |
| `embeddings.ts` | Pengendali penjanaan benam: menyelesaikan model pembenaman → pembekal, menghantar kepada API pembekal, mengembalikan respons pembenaman serasi OpenAI. Menyokong 6+ pembekal. |
| `imageGeneration.ts` | Pengendali penjanaan imej: menyelesaikan model imej → pembekal, menyokong mod serasi OpenAI, imej Gemini (Antigraviti) dan sandaran (Nebius). Mengembalikan imej base64 atau URL. |
| File | Purpose |
| --------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `chatCore.ts` | **Central orchestrator** (~600 lines). Handles the complete request lifecycle: format detection → translation → executor dispatch → streaming/non-streaming response → token refresh → error handling → usage logging. |
| `responsesHandler.ts` | Adapter for OpenAI's Responses API: converts Responses format → Chat Completions → sends to `chatCore` → converts SSE back to Responses format. |
| `embeddings.ts` | Embedding generation handler: resolves embedding model → provider, dispatches to provider API, returns OpenAI-compatible embedding response. Supports 6+ providers. |
| `imageGeneration.ts` | Image generation handler: resolves image model → provider, supports OpenAI-compatible, Gemini-image (Antigravity), and fallback (Nebius) modes. Returns base64 or URL images. |
#### Minta Kitaran Hayat (chatCore.ts)
#### Request Lifecycle (chatCore.ts)
```mermaid
sequenceDiagram
@ -258,28 +258,28 @@ sequenceDiagram
---
### 4.4 Perkhidmatan (`open-sse/services/`)
### 4.4 Services (`open-sse/services/`)
Logik perniagaan yang menyokong pengendali dan pelaksana.
Business logic that supports the handlers and executors.
| Fail | Tujuan |
| -------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `provider.ts` | **Pengesanan format** (`detectFormat`): menganalisis struktur badan permintaan untuk mengenal pasti format Claude/OpenAI/Gemini/Antigravity/Respons (termasuk `max_tokens` heuristik untuk Claude). Juga: Pembinaan URL, pembinaan pengepala, penormalan konfigurasi pemikiran. Menyokong `openai-compatible-*` dan `anthropic-compatible-*` pembekal dinamik. |
| `model.ts` | Penghuraian rentetan model (`claude/model-name``{provider: "claude", model: "model-name"}`), resolusi alias dengan pengesanan perlanggaran, pembersihan input (menolak aksara traversal/kawalan laluan) dan resolusi maklumat model dengan sokongan alias getter async. |
| `accountFallback.ts` | Pengendalian had kadar: pengunduran eksponen (1s → 2s → 4s → maks 2minit), pengurusan cooldown akaun, klasifikasi ralat (ralat yang mencetuskan sandaran berbanding tidak). |
| `tokenRefresh.ts` | Muat semula token OAuth untuk **setiap pembekal**: Google (Gemini, Antigravity), Claude, Codex, Qwen, iFlow, GitHub (OAuth + Copilot dwi-token), Kiro (AWS SSO OIDC + Social Auth). Termasuk cache penyahduplikasi janji dalam penerbangan dan cuba semula dengan pengunduran eksponen. |
| `combo.ts` | **Model kombo**: rangkaian model sandaran. Jika model A gagal dengan ralat layak sandar, cuba model B, kemudian C, dsb. Mengembalikan kod status huluan sebenar. |
| `usage.ts` | Mengambil data kuota/penggunaan daripada API pembekal (kuota Copilot GitHub, kuota model Antigraviti, had kadar Codex, pecahan penggunaan Kiro, tetapan Claude). |
| `accountSelector.ts` | Pemilihan akaun pintar dengan algoritma pemarkahan: mempertimbangkan keutamaan, status kesihatan, kedudukan round-robin dan keadaan cooldown untuk memilih akaun yang optimum bagi setiap permintaan. |
| `contextManager.ts` | Meminta pengurusan kitaran hayat konteks: mencipta dan menjejak objek konteks setiap permintaan dengan metadata (ID permintaan, cap masa, maklumat pembekal) untuk penyahpepijatan dan pengelogan. |
| `ipFilter.ts` | Kawalan capaian berasaskan IP: menyokong mod senarai dibenarkan dan senarai sekat. Mengesahkan IP klien terhadap peraturan yang dikonfigurasikan sebelum memproses permintaan API. |
| `sessionManager.ts` | Penjejakan sesi dengan cap jari pelanggan: menjejaki sesi aktif menggunakan pengecam pelanggan dicincang, memantau kiraan permintaan dan menyediakan metrik sesi. |
| `signatureCache.ts` | Minta cache penyahduplikasian berasaskan tandatangan: menghalang permintaan pendua dengan menyimpan cache tandatangan permintaan terkini dan mengembalikan respons cache untuk permintaan yang sama dalam tetingkap masa. |
| `systemPrompt.ts` | Suntikan gesaan sistem global: menambah atau menambahkan gesaan sistem yang boleh dikonfigurasikan kepada semua permintaan, dengan pengendalian keserasian setiap pembekal. |
| `thinkingBudget.ts` | Pengurusan belanjawan token penaakulan: menyokong mod laluan lalu, auto (konfigurasi pemikiran jalur), tersuai (belanjawan tetap) dan mod penyesuaian (berskala kerumitan) untuk mengawal token pemikiran/penaakulan. |
| `wildcardRouter.ts` | Penghalaan corak model kad liar: menyelesaikan corak kad bebas (cth., `*/claude-*`) kepada pasangan pembekal/model konkrit berdasarkan ketersediaan dan keutamaan. |
| File | Purpose |
| -------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `provider.ts` | **Format detection** (`detectFormat`): analyzes request body structure to identify Claude/OpenAI/Gemini/Antigravity/Responses formats (includes `max_tokens` heuristic for Claude). Also: URL building, header building, thinking config normalization. Supports `openai-compatible-*` and `anthropic-compatible-*` dynamic providers. |
| `model.ts` | Model string parsing (`claude/model-name``{provider: "claude", model: "model-name"}`), alias resolution with collision detection, input sanitization (rejects path traversal/control chars), and model info resolution with async alias getter support. |
| `accountFallback.ts` | Rate-limit handling: exponential backoff (1s → 2s → 4s → max 2min), account cooldown management, error classification (which errors trigger fallback vs. not). |
| `tokenRefresh.ts` | OAuth token refresh for **every provider**: Google (Gemini, Antigravity), Claude, Codex, Qwen, iFlow, GitHub (OAuth + Copilot dual-token), Kiro (AWS SSO OIDC + Social Auth). Includes in-flight promise deduplication cache and retry with exponential backoff. |
| `combo.ts` | **Combo models**: chains of fallback models. If model A fails with a fallback-eligible error, try model B, then C, etc. Returns actual upstream status codes. |
| `usage.ts` | Fetches quota/usage data from provider APIs (GitHub Copilot quotas, Antigravity model quotas, Codex rate limits, Kiro usage breakdowns, Claude settings). |
| `accountSelector.ts` | Smart account selection with scoring algorithm: considers priority, health status, round-robin position, and cooldown state to pick the optimal account for each request. |
| `contextManager.ts` | Request context lifecycle management: creates and tracks per-request context objects with metadata (request ID, timestamps, provider info) for debugging and logging. |
| `ipFilter.ts` | IP-based access control: supports allowlist and blocklist modes. Validates client IP against configured rules before processing API requests. |
| `sessionManager.ts` | Session tracking with client fingerprinting: tracks active sessions using hashed client identifiers, monitors request counts, and provides session metrics. |
| `signatureCache.ts` | Request signature-based deduplication cache: prevents duplicate requests by caching recent request signatures and returning cached responses for identical requests within a time window. |
| `systemPrompt.ts` | Global system prompt injection: prepends or appends a configurable system prompt to all requests, with per-provider compatibility handling. |
| `thinkingBudget.ts` | Reasoning token budget management: supports passthrough, auto (strip thinking config), custom (fixed budget), and adaptive (complexity-scaled) modes for controlling thinking/reasoning tokens. |
| `wildcardRouter.ts` | Wildcard model pattern routing: resolves wildcard patterns (e.g., `*/claude-*`) to concrete provider/model pairs based on availability and priority. |
#### Deduplikasi Token Refresh
#### Token Refresh Deduplication
```mermaid
sequenceDiagram
@ -300,7 +300,7 @@ sequenceDiagram
Cache->>Cache: Delete cache entry
```
#### Mesin Keadaan Fallback Akaun
#### Account Fallback State Machine
```mermaid
stateDiagram-v2
@ -325,7 +325,7 @@ stateDiagram-v2
}
```
#### Rantai Model Kombo
#### Combo Model Chain
```mermaid
flowchart LR
@ -344,11 +344,11 @@ flowchart LR
---
### 4.5 Penterjemah (`open-sse/translator/`)
### 4.5 Translator (`open-sse/translator/`)
**enjin terjemahan format** menggunakan sistem pemalam pendaftaran sendiri.
The **format translation engine** using a self-registering plugin system.
#### Seni bina
#### Architecture
```mermaid
graph TD
@ -374,15 +374,15 @@ graph TD
end
```
| Direktori | Fail | Penerangan |
| ------------ | ------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `request/` | 8 penterjemah | Tukar badan permintaan antara format. Setiap fail mendaftar sendiri melalui `register(from, to, fn)` semasa diimport. |
| `response/` | 7 penterjemah | Tukar ketulan respons penstriman antara format. Mengendalikan jenis acara SSE, blok pemikiran, panggilan alat. |
| `helpers/` | 6 pembantu | Utiliti dikongsi: `claudeHelper` (pengekstrak segera sistem, konfigurasi pemikiran), `geminiHelper` (pemetaan bahagian/kandungan), `openaiHelper` (penapisan format), `toolCallHelper` (penjanaan ID, suntikan tindak balas tiada), `toolCallHelper`, `toolCallHelper` |
| `index.ts` | — | Enjin terjemahan: `translateRequest()`, `translateResponse()`, pengurusan negeri, pendaftaran. |
| `formats.ts` | — | Pemalar format: `OPENAI`, `CLAUDE`, `GEMINI`, `ANTIGRAVITY`, `KIRO`, `CURSOR`, `OPENAI_RESPONSES`. |
| Directory | Files | Description |
| ------------ | ------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `request/` | 8 translators | Convert request bodies between formats. Each file self-registers via `register(from, to, fn)` on import. |
| `response/` | 7 translators | Convert streaming response chunks between formats. Handles SSE event types, thinking blocks, tool calls. |
| `helpers/` | 6 helpers | Shared utilities: `claudeHelper` (system prompt extraction, thinking config), `geminiHelper` (parts/contents mapping), `openaiHelper` (format filtering), `toolCallHelper` (ID generation, missing response injection), `maxTokensHelper`, `responsesApiHelper`. |
| `index.ts` | — | Translation engine: `translateRequest()`, `translateResponse()`, state management, registry. |
| `formats.ts` | — | Format constants: `OPENAI`, `CLAUDE`, `GEMINI`, `ANTIGRAVITY`, `KIRO`, `CURSOR`, `OPENAI_RESPONSES`. |
#### Reka Bentuk Utama: Pemalam Mendaftar Sendiri
#### Key Design: Self-Registering Plugins
```javascript
// Each translator file calls register() on import:
@ -395,19 +395,19 @@ import "./request/claude-to-openai.js"; // ← self-registers
---
### 4.6 Util (`open-sse/utils/`)
### 4.6 Utils (`open-sse/utils/`)
| Fail | Tujuan |
| ------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `error.ts` | Pembinaan tindak balas ralat (format serasi OpenAI), penghuraian ralat huluan, Pengekstrakan masa percubaan semula Antigraviti daripada mesej ralat, penstriman ralat SSE. |
| `stream.ts` | **SSE Transform Stream**saluran paip penstriman teras. Dua mod: `TRANSLATE` (terjemahan format penuh) dan `PASSTHROUGH` (normalkan + penggunaan ekstrak). Mengendalikan penimbalan bongkah, anggaran penggunaan, penjejakan panjang kandungan. Kejadian pengekod/penyahkod setiap aliran mengelakkan keadaan dikongsi. |
| `streamHelpers.ts` | Utiliti SSE peringkat rendah: `parseSSELine` (bertoleransi ruang putih), `hasValuableContent` (menapis ketulan kosong untuk OpenAI/Claude/Gemini), `fixInvalidId`, `formatSSE` (serialisasi SSE sedar format dengan pembersihan `perf_metrics`). |
| `usageTracking.ts` | Pengekstrakan penggunaan token daripada sebarang format (Claude/OpenAI/Gemini/Responses), anggaran dengan nisbah char-per-token alat/mesej yang berasingan, penambahan penimbal (margin keselamatan 2000 token), penapisan medan khusus format, pengelogan konsol dengan warna ANSI. |
| `requestLogger.ts` | Pengelogan permintaan berasaskan fail (ikut serta melalui `ENABLE_REQUEST_LOGS=true`). Mencipta folder sesi dengan fail bernombor: `1_req_client.json``7_res_client.txt`. Semua I/O tidak segerak (api-dan-lupa). Topeng tajuk sensitif. |
| `bypassHandler.ts` | Memintas corak tertentu daripada Claude CLI (pengeluaran tajuk, pemanasan, kiraan) dan mengembalikan respons palsu tanpa menghubungi mana-mana pembekal. Menyokong kedua-dua penstriman dan bukan penstriman. Sengaja dihadkan kepada skop Claude CLI. |
| `networkProxy.ts` | Menyelesaikan URL proksi keluar untuk pembekal tertentu dengan keutamaan: konfigurasi khusus pembekal → konfigurasi global → pembolehubah persekitaran (`HTTPS_PROXY`/`HTTP_PROXY`/`ALL_PROXY`). Menyokong `NO_PROXY` pengecualian. Konfigurasi cache untuk 30s. |
| File | Purpose |
| ------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| `error.ts` | Error response building (OpenAI-compatible format), upstream error parsing, Antigravity retry-time extraction from error messages, SSE error streaming. |
| `stream.ts` | **SSE Transform Stream**the core streaming pipeline. Two modes: `TRANSLATE` (full format translation) and `PASSTHROUGH` (normalize + extract usage). Handles chunk buffering, usage estimation, content length tracking. Per-stream encoder/decoder instances avoid shared state. |
| `streamHelpers.ts` | Low-level SSE utilities: `parseSSELine` (whitespace-tolerant), `hasValuableContent` (filters empty chunks for OpenAI/Claude/Gemini), `fixInvalidId`, `formatSSE` (format-aware SSE serialization with `perf_metrics` cleanup). |
| `usageTracking.ts` | Token usage extraction from any format (Claude/OpenAI/Gemini/Responses), estimation with separate tool/message char-per-token ratios, buffer addition (2000 tokens safety margin), format-specific field filtering, console logging with ANSI colors. |
| `requestLogger.ts` | File-based request logging (opt-in via `ENABLE_REQUEST_LOGS=true`). Creates session folders with numbered files: `1_req_client.json``7_res_client.txt`. All I/O is async (fire-and-forget). Masks sensitive headers. |
| `bypassHandler.ts` | Intercepts specific patterns from Claude CLI (title extraction, warmup, count) and returns fake responses without calling any provider. Supports both streaming and non-streaming. Intentionally limited to Claude CLI scope. |
| `networkProxy.ts` | Resolves outbound proxy URL for a given provider with precedence: provider-specific config → global config → environment variables (`HTTPS_PROXY`/`HTTP_PROXY`/`ALL_PROXY`). Supports `NO_PROXY` exclusions. Caches config for 30s. |
#### Saluran Paip Penstriman SSE
#### SSE Streaming Pipeline
```mermaid
flowchart TD
@ -429,7 +429,7 @@ flowchart TD
style M fill:#9f9,stroke:#333
```
#### Permintaan Struktur Sesi Logger
#### Request Logger Session Structure
```
logs/
@ -447,109 +447,109 @@ logs/
---
### 4.7 Lapisan Aplikasi (`src/`)
### 4.7 Application Layer (`src/`)
| Direktori | Tujuan |
| ------------- | ----------------------------------------------------------------------------- |
| `src/app/` | UI Web, laluan API, perisian tengah Ekspres, pengendali panggil balik OAuth |
| `src/lib/` | Akses pangkalan data (`localDb.ts`, `usageDb.ts`), pengesahan, dikongsi |
| `src/mitm/` | Utiliti proksi man-in-the-middle untuk memintas trafik pembekal |
| `src/models/` | Takrif model pangkalan data |
| `src/shared/` | Pembalut di sekeliling fungsi open-sse (penyedia, strim, ralat, dll.) |
| `src/sse/` | Pengendali titik akhir SSE yang menghantar pustaka open-sse ke laluan Express |
| `src/store/` | Pengurusan keadaan aplikasi |
| Directory | Purpose |
| ------------- | ---------------------------------------------------------------------- |
| `src/app/` | Web UI, API routes, Express middleware, OAuth callback handlers |
| `src/lib/` | Database access (`localDb.ts`, `usageDb.ts`), authentication, shared |
| `src/mitm/` | Man-in-the-middle proxy utilities for intercepting provider traffic |
| `src/models/` | Database model definitions |
| `src/shared/` | Wrappers around open-sse functions (provider, stream, error, etc.) |
| `src/sse/` | SSE endpoint handlers that wire the open-sse library to Express routes |
| `src/store/` | Application state management |
#### Laluan API Terkenal
#### Notable API Routes
| Laluan | Kaedah | Tujuan |
| --------------------------------------------- | -------------------- | ----------------------------------------------------------------------------------------- |
| `/api/provider-models` | DAPATKAN/POST/PADAM | CRUD untuk model tersuai setiap pembekal |
| `/api/models/catalog` | DAPATKAN | Katalog agregat semua model (sembang, benam, imej, tersuai) dikumpulkan mengikut pembekal |
| `/api/settings/proxy` | DAPATKAN/LETAK/PADAM | Konfigurasi proksi keluar hierarki (`global/providers/combos/keys`) |
| `/api/settings/proxy/test` | POS | Mengesahkan sambungan proksi dan mengembalikan IP/kependaman awam |
| `/v1/providers/[provider]/chat/completions` | POS | Penyelesaian sembang khusus bagi setiap pembekal dengan pengesahan model |
| `/v1/providers/[provider]/embeddings` | POS | Pembenaman khusus bagi setiap pembekal dengan pengesahan model |
| `/v1/providers/[provider]/images/generations` | POS | Penjanaan imej setiap pembekal khusus dengan pengesahan model |
| `/api/settings/ip-filter` | DAPATKAN/LETAK | Pengurusan senarai dibenarkan/senarai sekat IP |
| `/api/settings/thinking-budget` | DAPATKAN/LETAK | Konfigurasi belanjawan token penaakulan (laluan/auto/tersuai/suai) |
| `/api/settings/system-prompt` | DAPATKAN/LETAK | Suntikan segera sistem global untuk semua permintaan |
| `/api/sessions` | DAPATKAN | Penjejakan dan metrik sesi aktif |
| `/api/rate-limits` | DAPATKAN | Status had kadar setiap akaun |
| Route | Methods | Purpose |
| --------------------------------------------- | --------------- | ------------------------------------------------------------------------------------- |
| `/api/provider-models` | GET/POST/DELETE | CRUD for custom models per provider |
| `/api/models/catalog` | GET | Aggregated catalog of all models (chat, embedding, image, custom) grouped by provider |
| `/api/settings/proxy` | GET/PUT/DELETE | Hierarchical outbound proxy configuration (`global/providers/combos/keys`) |
| `/api/settings/proxy/test` | POST | Validates proxy connectivity and returns public IP/latency |
| `/v1/providers/[provider]/chat/completions` | POST | Dedicated per-provider chat completions with model validation |
| `/v1/providers/[provider]/embeddings` | POST | Dedicated per-provider embeddings with model validation |
| `/v1/providers/[provider]/images/generations` | POST | Dedicated per-provider image generation with model validation |
| `/api/settings/ip-filter` | GET/PUT | IP allowlist/blocklist management |
| `/api/settings/thinking-budget` | GET/PUT | Reasoning token budget configuration (passthrough/auto/custom/adaptive) |
| `/api/settings/system-prompt` | GET/PUT | Global system prompt injection for all requests |
| `/api/sessions` | GET | Active session tracking and metrics |
| `/api/rate-limits` | GET | Per-account rate limit status |
---
## 5. Corak Reka Bentuk Utama
## 5. Key Design Patterns
### 5.1 Terjemahan Hub-and-Spoke
### 5.1 Hub-and-Spoke Translation
Semua format diterjemahkan melalui **format OpenAI sebagai hab**. Menambah penyedia baharu hanya memerlukan penulisan **sepasang** penterjemah (ke/dari OpenAI), bukan N pasangan.
All formats translate through **OpenAI format as the hub**. Adding a new provider only requires writing **one pair** of translators (to/from OpenAI), not N pairs.
### 5.2 Corak Strategi Pelaksana
### 5.2 Executor Strategy Pattern
Setiap pembekal mempunyai kelas pelaksana khusus yang diwarisi daripada `BaseExecutor`. Kilang di `executors/index.ts` memilih yang betul semasa masa jalan.
Each provider has a dedicated executor class inheriting from `BaseExecutor`. The factory in `executors/index.ts` selects the right one at runtime.
### 5.3 Sistem Pemalam Mendaftar Sendiri
### 5.3 Self-Registering Plugin System
Modul penterjemah mendaftarkan diri mereka pada import melalui `register()`. Menambah penterjemah baharu hanyalah mencipta fail dan mengimportnya.
Translator modules register themselves on import via `register()`. Adding a new translator is just creating a file and importing it.
### 5.4 Pengunduran Akaun dengan Pengunduran Eksponen
### 5.4 Account Fallback with Exponential Backoff
Apabila pembekal mengembalikan 429/401/500, sistem boleh bertukar ke akaun seterusnya, menggunakan tempoh bertenang eksponen (1s → 2s → 4s → maks 2min).
When a provider returns 429/401/500, the system can switch to the next account, applying exponential cooldowns (1s → 2s → 4s → max 2min).
### 5.5 Rantai Model Kombo
### 5.5 Combo Model Chains
"Kombo" mengumpulkan berbilang rentetan `provider/model`. Jika yang pertama gagal, sandarkan kepada yang seterusnya secara automatik.
A "combo" groups multiple `provider/model` strings. If the first fails, fallback to the next automatically.
### 5.6 Terjemahan Penstriman Stateful
### 5.6 Stateful Streaming Translation
Terjemahan respons mengekalkan keadaan merentas bahagian SSE (penjejakan blok pemikiran, pengumpulan panggilan alat, pengindeksan blok kandungan) melalui mekanisme `initState()`.
Response translation maintains state across SSE chunks (thinking block tracking, tool call accumulation, content block indexing) via the `initState()` mechanism.
### 5.7 Penimbal Keselamatan Penggunaan
### 5.7 Usage Safety Buffer
Penampan 2000-token ditambahkan pada penggunaan yang dilaporkan untuk menghalang pelanggan daripada mencapai had tetingkap konteks kerana overhed daripada gesaan sistem dan terjemahan format.
A 2000-token buffer is added to reported usage to prevent clients from hitting context window limits due to overhead from system prompts and format translation.
---
## 6. Format yang Disokong
## 6. Supported Formats
| Format | Arah | Pengecam |
| ---------------------- | ---------------- | ------------------ |
| Selesai Sembang OpenAI | sumber + sasaran | `openai` |
| API Respons OpenAI | sumber + sasaran | `openai-responses` |
| Claude Anthropic | sumber + sasaran | `claude` |
| Google Gemini | sumber + sasaran | `gemini` |
| Google Gemini CLI | sasaran sahaja | `gemini-cli` |
| Antigraviti | sumber + sasaran | `antigravity` |
| AWS Kiro | sasaran sahaja | `kiro` |
| Kursor | sasaran sahaja | `cursor` |
| Format | Direction | Identifier |
| ----------------------- | --------------- | ------------------ |
| OpenAI Chat Completions | source + target | `openai` |
| OpenAI Responses API | source + target | `openai-responses` |
| Anthropic Claude | source + target | `claude` |
| Google Gemini | source + target | `gemini` |
| Google Gemini CLI | target only | `gemini-cli` |
| Antigravity | source + target | `antigravity` |
| AWS Kiro | target only | `kiro` |
| Cursor | target only | `cursor` |
---
## 7. Pembekal yang Disokong
## 7. Supported Providers
| Pembekal | Kaedah Pengesahan | Pelaksana | Nota Utama |
| ------------------------ | ------------------------ | ----------- | -------------------------------------------------------- |
| Claude Anthropic | Kunci API atau OAuth | Lalai | Menggunakan pengepala `x-api-key` |
| Google Gemini | Kunci API atau OAuth | Lalai | Menggunakan pengepala `x-goog-api-key` |
| Google Gemini CLI | OAuth | GeminiCLI | Menggunakan `streamGenerateContent` titik akhir |
| Antigraviti | OAuth | Antigraviti | Undur berbilang URL, penghuraian cuba semula tersuai |
| OpenAI | Kunci API | Lalai | Pengesahan Pembawa Standard |
| Codex | OAuth | Codex | Menyuntik arahan sistem, mengurus pemikiran |
| GitHub Copilot | Token OAuth + Copilot | Github | Token dwi, pengepala VSCode meniru |
| Kiro (AWS) | AWS SSO OIDC atau Sosial | Kiro | Perduaan EventStream parsing |
| IDE kursor | Pengesahan semak | Kursor | Pengekodan Protobuf, jumlah semak SHA-256 |
| Qwen | OAuth | Lalai | Pengesahan standard |
| iFlow | OAuth (Asas + Pembawa) | Lalai | Pengepala dwi pengesahan |
| OpenRouter | Kunci API | Lalai | Pengesahan Pembawa Standard |
| GLM, Kimi, MiniMax | Kunci API | Lalai | Serasi Claude, gunakan `x-api-key` |
| `openai-compatible-*` | Kunci API | Lalai | Dinamik: mana-mana titik akhir serasi OpenAI |
| `anthropic-compatible-*` | Kunci API | Lalai | Dinamik: mana-mana titik akhir yang serasi dengan Claude |
| Provider | Auth Method | Executor | Key Notes |
| ------------------------ | ---------------------- | ----------- | --------------------------------------------- |
| Anthropic Claude | API key or OAuth | Default | Uses `x-api-key` header |
| Google Gemini | API key or OAuth | Default | Uses `x-goog-api-key` header |
| Google Gemini CLI | OAuth | GeminiCLI | Uses `streamGenerateContent` endpoint |
| Antigravity | OAuth | Antigravity | Multi-URL fallback, custom retry parsing |
| OpenAI | API key | Default | Standard Bearer auth |
| Codex | OAuth | Codex | Injects system instructions, manages thinking |
| GitHub Copilot | OAuth + Copilot token | Github | Dual token, VSCode header mimicking |
| Kiro (AWS) | AWS SSO OIDC or Social | Kiro | Binary EventStream parsing |
| Cursor IDE | Checksum auth | Cursor | Protobuf encoding, SHA-256 checksums |
| Qwen | OAuth | Default | Standard auth |
| iFlow | OAuth (Basic + Bearer) | Default | Dual auth header |
| OpenRouter | API key | Default | Standard Bearer auth |
| GLM, Kimi, MiniMax | API key | Default | Claude-compatible, use `x-api-key` |
| `openai-compatible-*` | API key | Default | Dynamic: any OpenAI-compatible endpoint |
| `anthropic-compatible-*` | API key | Default | Dynamic: any Claude-compatible endpoint |
---
## 8. Ringkasan Aliran Data
## 8. Data Flow Summary
### Permintaan Penstriman
### Streaming Request
```mermaid
flowchart LR
@ -566,7 +566,7 @@ flowchart LR
K --> L["logUsage()\nsaveRequestUsage()"]
```
### Permintaan Bukan Penstriman
### Non-Streaming Request
```mermaid
flowchart LR
@ -577,7 +577,7 @@ flowchart LR
E --> F["Return JSON\nresponse"]
```
### Aliran Pintasan (Claude CLI)
### Bypass Flow (Claude CLI)
```mermaid
flowchart LR