Skip to content

Commit 07bd538

Browse files
longsizhuolongsizhuo
committed
feat(docs): i18n computer-science + jobs + all-projects 翻译完成 (23 篇)
translator-cs-jobs 产出： - computer-science: 双向翻译 - data-structures (en→zh, 5 篇): index / array / linked-list - frontend (zh→en, 2 篇) - cpp_backend (zh→en, 8 篇): mempool / threadpool / 编译系列 - index.mdx (zh→en) - jobs (zh→en, 5 篇): event-keynote 2 / interview-prep 3 - all-projects (zh→en, 2 篇): ai-town / multimodal-rl 代码块内变量名和 API 原样保留，仅翻译注释。 frontmatter 继承原文 docId，带 translatedFrom 标记。
1 parent 7a15e9b commit 07bd538

23 files changed

Lines changed: 3131 additions & 0 deletions

‎app/docs/all-projects/ai-town.en.mdx‎

Lines changed: 74 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,74 @@
1+
---
2+
title: AI Town Design Document
3+
description: ""
4+
date: "2025-10-18"
5+
tags:
6+
- ai-project
7+
docId: bkxwg1m9p9rnm8062wsm020w
8+
lang: en
9+
translatedFrom: zh
10+
translatedAt: 2026-04-15T08:00:00Z
11+
translatorAgent: claude-sonnet-4-6
12+
---
13+
14+
# AI Town Design Document
15+
16+
## 1. Project Overview
17+
18+
- **Type**: A lightweight simulation + social + quest-based mini-game driven by multi-agent NPCs
19+
- **Core selling points**: NPCs "remember you" and collaborate with each other through dialogue; players can use **community contribution points** (earned by posting, submitting PRs, etc.) to obtain in-game currency and abilities, driving town events
20+
- **Technology foundation**: Godot 4 (Microverse-style) + multi-agent framework (O-R-P-A: Observe → Retrieve → Plan → Act) + local model first (with template fallback)
21+
22+
## 2. Goals (MVP)
23+
24+
1. Single map + 3 NPCs (merchant / messenger / editor) + quest board (fetch / relay / check-in)
25+
2. Dialogue with **short-term memory + end-of-day summary**
26+
3. **Minimal community integration**: support entering a "redemption code" to receive coins / action points (future: automatic issuance via webhook)
27+
4. Use points (or redeemed coins) to trigger 2–3 **visible world changes** (discount day / extra quests / expanded dialogue budget)
28+
29+
## 3. Core Gameplay (Version 1)
30+
31+
- **Loop**: Accept quest → Dialogue / collaborate with NPCs → Complete to earn coins / AP → Nightly summary generated → Events refresh next day
32+
- **Uses of points / coins** (pick 2–3 to implement first)
33+
- Unlock a **discount day** at the shop (all prices -10%)
34+
- Purchase **action points** (one extra quest per day)
35+
- Purchase **dialogue budget** (3 additional conversation turns with an NPC that day)
36+
- Trigger a **theme-day announcement** (published by the editor NPC; NPC dialogue becomes more active)
37+
38+
## 4. Open-Source Community Integration (Two Phases)
39+
40+
### Phase A (MVP) — Redemption Code Verification
41+
42+
- The community backend issues one-time **redemption codes** (containing point value and expiry); player enters code in-game → server verifies and voids → returns coins / AP
43+
- **Advantage**: No login or account binding required; maximally stable and ready to ship
44+
45+
### Phase B (Mid-term) — Automatic Issuance via Webhook
46+
47+
- Posts, PR merges on GitHub / the site trigger a Webhook → write to `pending_rewards`
48+
- Game launch or clicking "Sync" → fetch pending rewards → automatically credited
49+
- Optional: bind Steam / GitHub account for stronger identity verification
50+
51+
## 5. System Architecture (Minimal Modules)
52+
53+
- **Client (Godot)**
54+
- `Wallet` (authoritative entry point for coins / AP)
55+
- `TaskManager`, `DialogManager`, `MemoryManager`, `CharacterManager`
56+
- `TownEventBus` (broadcasts shop open / midday break / close / theme day)
57+
- `RedeemPanel` (redemption code UI)
58+
59+
- **Services (can be merged into community backend)**
60+
- `/api/v1/redeem` (one-time verification and voiding)
61+
- (Reserved) `/api/v1/rewards/pending`, `/webhooks/github`
62+
63+
**Data Flow (MVP)**
64+
Community issues code → Player enters it in-game → `redeem` verifies → Returns coins / AP → `Wallet` credits → `TownEventBus` triggers discount / quest refresh
65+
66+
## 6. Scoring and Spending (Initial Draft)
67+
68+
| Action | Community Points Earned | In-game Conversion (Example) |
69+
| ---------------- | ----------------------: | ---------------------------- |
70+
| Post approved | +80 | 80 pts = 400 coins |
71+
| PR merged | +80 | 80 pts = 400 coins |
72+
| Article featured | +50 | 50 pts = 1 "theme day" item |
73+
74+
> Conversion rates are stored in a config file; events can apply temporary bonuses (e.g., 1.2× on weekends)

‎app/docs/all-projects/multimodal-rl.en.mdx‎

Lines changed: 94 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,94 @@
1+
---
2+
title: Multimodal Reinforcement Learning Project (MVP Goals)
3+
description: Build a lightweight multimodal understanding and generation system that closes the loop from visual perception to language expression, incorporating reinforcement learning and answer-to-image generation.
4+
date: "2025-10-17"
5+
tags:
6+
- projects
7+
- multimodal
8+
- reinforcement-learning
9+
- RLHF
10+
docId: ifwz8sqxqsgjrafa79pycrcm
11+
lang: en
12+
translatedFrom: zh
13+
translatedAt: 2026-04-15T08:00:00Z
14+
translatorAgent: claude-sonnet-4-6
15+
---
16+
17+
# Multimodal Group – MVP Specification
18+
19+
**Project version:** v0.1
20+
**Repository:** [involutionhell](https://github.com/InvolutionHell/involutionhell)
21+
22+
---
23+
24+
<a id="vision"></a>
25+
## 1. Vision
26+
27+
Build a lightweight multimodal understanding and generation system that enables the model to interpret images, retrieve relevant information, and produce logically coherent text output.
28+
The goal is to close the full loop from visual perception to language expression, and further develop the ability to explain answers through generated images.
29+
30+
<a id="mvp-goals"></a>
31+
## 2. MVP Phase Goals
32+
33+
<a id="phase-1"></a>
34+
### Phase 1: Basic Multimodal Pipeline
35+
36+
- Image content recognition (objects, scenes, semantic labels).
37+
- Semantic retrieval (image → text / text → image).
38+
- Generative understanding and text output.
39+
- Model references: CLIP / SigLIP / BLIP-2 / LLaVA / Qwen-VL.
40+
41+
<a id="phase-2"></a>
42+
### Phase 2: Multimodal Reinforcement Learning
43+
44+
- Incorporate user feedback and reward signals to optimise model generation and retrieval performance.
45+
- Main directions:
46+
1. RLHF / DPO fine-tuning to learn user preferences.
47+
2. Retrieval strategy optimisation based on behavioural data.
48+
3. Generation quality control and consistency improvement.
49+
50+
- Goal: give the system the ability to self-improve and adapt to user preferences.
51+
52+
<a id="phase-2-5"></a>
53+
### Phase 2.5: Answer-to-Image Generation
54+
55+
- Automatically generate illustrative images from the model's text answers to aid comprehension.
56+
- Implementation: use Stable Diffusion / SDXL to convert answer text into image prompts.
57+
- Application examples:
58+
- Answer "the process of black hole formation" → generate a structural diagram.
59+
- Explain a scene from a novel → generate a conceptual illustration.
60+
61+
- Goal: enable the system not only to understand images and answer questions, but also to explain answers through generated images.
62+
63+
<a id="architecture"></a>
64+
## 3. System Architecture
65+
66+
```
67+
[Frontend] → Upload image / Display results
68+
↓
69+
[Backend API] → FastAPI + LangChain + Vector Search
70+
↓
71+
[Multimodal Models] → CLIP / BLIP / LLaVA / Qwen-VL
72+
↓
73+
[RL Module + Answer-to-Image] (Phase 2 and 2.5)
74+
```
75+
76+
<a id="milestones"></a>
77+
## 4. Milestones
78+
79+
| Phase | Goal | Deliverables |
80+
| --------- | ------------------------------------- | --------------------------------------------- |
81+
| Phase 1 | Multimodal recognition and generation | Image recognition, retrieval, text generation |
82+
| Phase 2 | Reinforcement learning optimisation | RLHF / DPO, retrieval strategy optimisation |
83+
| Phase 2.5 | Answer-to-image generation | Automatic illustration generation |
84+
| Phase 3 | Scaling and deployment | Web demo and API interface |
85+
86+
<a id="team"></a>
87+
## 5. Team Responsibilities
88+
89+
| Module | Owner |
90+
| ----------------------------------------------- | -------- |
91+
| Image recognition and encoding | Member A |
92+
| Semantic retrieval and data processing | Member B |
93+
| Generation module and model integration | Member C |
94+
| Reinforcement learning and visualisation output | Member D |

0 commit comments

Comments
 (0)