-
Install dependencies: Make sure you have Bun installed, then run:
bun install
-
Build the UI for production:
cd src/web && bun run ui:build
-
Copy and paste the
.env.example
and populate the environment variables -
Start the server:
bun run server
This plugin revolutionizes open source collaboration by implementing an AI-powered reward mechanism for quality contributions.
At the heart of the system is a content evaluation module that assigns monetary value to contributor comments in the context of work projects with specifications. Here's how it works:
-
The system processes both issue comments and pull request review comments through different evaluation pipelines, with comprehensive preprocessing that:
- Removes user commands (starting with /) and bot responses
- Filters out quoted text (starting with >)
- Removes HTML comments and footnotes
- For assigned users, considers comment timestamps to optionally exclude those posted during assignment periods, to reduce gaming
- Processes linked pull request comments through GraphQL API
- Handles minimized/hidden comments
- Credits only unique links to prevent duplicates
-
For issue comments, it generates a context-aware prompt that includes:
- The original issue description and specification
- All comments in the conversation for context
- The specific comments being evaluated
-
The evaluation process handles GitHub-flavored markdown intelligently:
- It distinguishes between quoted text (starting with '>') and original content
- Only evaluates the commenter's original contributions
- Considers the relationship between comments and their context
-
The language model assigns relevance scores from 0 to 1:
interface Relevances { [commentId: string]: number; // 0 = irrelevant, 1 = highly relevant }
The review incentivization module implements a sophisticated algorithm for rewarding code reviews:
interface ReviewScore {
reviewId: number;
effect: {
addition: number;
deletion: number;
};
reward: number;
priority: number;
}
The system calculates rewards based on:
- The scope of code reviewed (additions + deletions)
- Issue priority labels
- The conclusiveness of the review (APPROVED or CHANGES_REQUESTED states receive additional credit)
- File-specific exclusions through pattern matching
The permit generation module handles the secure distribution of rewards:
-
Security Checks:
- Validates that the issue is collaborative
- Verifies private key permissions against organization and repository IDs
- Implements a multi-format encryption system for private keys
-
Fee Processing:
- Automatically calculates and deducts platform fees
- Supports token-specific fee exemptions through whitelist
- Creates treasury allocations for fee distribution
-
Reward Distribution:
- Generates ERC20 token permits for each contributor
- Stores permit data securely in a Supabase database
- Creates claimable reward URLs in the format:
https://pay.ubq.fi?claim=[encoded_permit]
The system uses decimal.js for precise token calculations:
const feeRateDecimal = new Decimal(100).minus(env.PERMIT_FEE_RATE).div(100);
const totalAfterFee = new Decimal(rewardResult.total).mul(feeRateDecimal).toNumber();
For large conversations, the system implements intelligent token management:
// Dynamically handles token limits and chunking for large conversations
_calculateMaxTokens(prompt: string, totalTokenLimit: number = 16384) {
// Token limit is configurable and adjusts based on model and rate limits
const inputTokens = this.tokenizer.encode(prompt).length;
const limit = Math.min(this._configuration?.tokenCountLimit, this._rateLimit);
return Math.min(inputTokens, limit);
}
// Splits large conversations into manageable chunks
async _splitPromptForEvaluation(specification: string, comments: Comment[]) {
let chunks = 2;
while (this._exceedsTokenLimit(comments, chunks)) {
chunks++;
}
return this._processChunks(specification, comments, chunks);
}
The system maintains a comprehensive record of all permits and rewards:
interface PermitRecord {
amount: string;
nonce: string;
deadline: string;
signature: string;
beneficiary_id: number;
location_id: number;
}
{
"userName": {
"comments": [
{
"content": "comment content",
"url": "https://url-to-item",
"type": 18,
"score": {
"formatting": {
"content": {
"p": {
"count": 16,
"score": 1
}
},
"wordValue": 0.1,
"multiplier": 1
},
"reward": 0.8,
"relevance": 0.5
}
}
],
"total": 40.5,
"task": {
"reward": 37.5,
"multiplier": 1
},
"feeRate": 0.1,
"permitUrl": "https://example.com/permit",
"userId": 123,
"evaluationCommentHtml": "<p>Evaluation comment</p>"
}
}
Reward formula:
Here is a possible valid configuration to enable this plugin. See these files for more details.
plugin: ubiquity-os/conversation-rewards
with:
evmNetworkId: 100
evmPrivateEncrypted: "encrypted-key"
erc20RewardToken: "0xe91D153E0b41518A2Ce8Dd3D7944Fa863463a97d"
dataCollection:
maxAttempts: 10
delayMs: 10000
incentives:
requirePriceLabel: true
limitRewards: true
collaboratorOnlyPaymentInvocation: true
contentEvaluator:
llm:
model: "gpt-4" # Model identifier
endpoint: "https://api.openrouter.ai/api/v1" # Configurable LLM endpoint
tokenCountLimit: 124000 # Adjustable token limit
maxRetries: 3 # Number of retries for rate limits/errors
multipliers:
- role: [ISSUE_SPECIFICATION]
relevance: 1
- role: [PULL_AUTHOR]
relevance: 1
- role: [PULL_ASSIGNEE]
relevance: 1
- role: [PULL_COLLABORATOR]
relevance: 1
- role: [PULL_CONTRIBUTOR]
relevance: 1
userExtractor:
redeemTask: true
dataPurge:
skipCommentsWhileAssigned: all
reviewIncentivizer:
baseRate: 100
conclusiveReviewCredit: 25
formattingEvaluator:
wordCountExponent: 0.85
multipliers:
- role: ["ISSUE_SPECIFICATION"]
multiplier: 1
rewards:
html:
br: { score: 0, countWords: true }
code: { score: 5, countWords: false }
p: { score: 0, countWords: true }
em: { score: 0, countWords: true }
img: { score: 5, countWords: true }
strong: { score: 0, countWords: false }
blockquote: { score: 0, countWords: false }
h1: { score: 1, countWords: true }
h2: { score: 1, countWords: true }
h3: { score: 1, countWords: true }
h4: { score: 1, countWords: true }
h5: { score: 1, countWords: true }
h6: { score: 1, countWords: true }
a: { score: 5, countWords: true }
li: { score: 0.5, countWords: true }
ul: { score: 1, countWords: true }
td: { score: 0, countWords: true }
hr: { score: 0, countWords: true }
pre: { score: 0, countWords: false }
ol: { score: 1, countWords: true }
wordValue: 0.1
- role: ["ISSUE_AUTHOR"]
multiplier: 1
rewards:
html:
br: { score: 0, countWords: true }
code: { score: 5, countWords: false }
p: { score: 0, countWords: true }
em: { score: 0, countWords: true }
img: { score: 5, countWords: true }
strong: { score: 0, countWords: false }
blockquote: { score: 0, countWords: false }
h1: { score: 1, countWords: true }
h2: { score: 1, countWords: true }
h3: { score: 1, countWords: true }
h4: { score: 1, countWords: true }
h5: { score: 1, countWords: true }
h6: { score: 1, countWords: true }
a: { score: 5, countWords: true }
li: { score: 0.5, countWords: true }
ul: { score: 1, countWords: true }
td: { score: 0, countWords: true }
hr: { score: 0, countWords: true }
pre: { score: 0, countWords: false }
ol: { score: 1, countWords: true }
wordValue: 0.2
- role: ["ISSUE_ASSIGNEE"]
multiplier: 0
rewards:
html:
br: { score: 0, countWords: true }
code: { score: 5, countWords: false }
p: { score: 0, countWords: true }
em: { score: 0, countWords: true }
img: { score: 5, countWords: true }
strong: { score: 0, countWords: false }
blockquote: { score: 0, countWords: false }
h1: { score: 1, countWords: true }
h2: { score: 1, countWords: true }
h3: { score: 1, countWords: true }
h4: { score: 1, countWords: true }
h5: { score: 1, countWords: true }
h6: { score: 1, countWords: true }
a: { score: 5, countWords: true }
li: { score: 0.5, countWords: true }
ul: { score: 1, countWords: true }
td: { score: 0, countWords: true }
hr: { score: 0, countWords: true }
pre: { score: 0, countWords: false }
ol: { score: 1, countWords: true }
wordValue: 0
- role: ["ISSUE_COLLABORATOR"]
multiplier: 1
rewards:
html:
br: { score: 0, countWords: true }
code: { score: 5, countWords: false }
p: { score: 0, countWords: true }
em: { score: 0, countWords: true }
img: { score: 5, countWords: true }
strong: { score: 0, countWords: false }
blockquote: { score: 0, countWords: false }
h1: { score: 1, countWords: true }
h2: { score: 1, countWords: true }
h3: { score: 1, countWords: true }
h4: { score: 1, countWords: true }
h5: { score: 1, countWords: true }
h6: { score: 1, countWords: true }
a: { score: 5, countWords: true }
li: { score: 0.5, countWords: true }
ul: { score: 1, countWords: true }
td: { score: 0, countWords: true }
hr: { score: 0, countWords: true }
pre: { score: 0, countWords: false }
ol: { score: 1, countWords: true }
wordValue: 0.1
- role: ["ISSUE_CONTRIBUTOR"]
multiplier: 0.25
rewards:
html:
br: { score: 0, countWords: true }
code: { score: 5, countWords: false }
p: { score: 0, countWords: true }
em: { score: 0, countWords: true }
img: { score: 5, countWords: true }
strong: { score: 0, countWords: false }
blockquote: { score: 0, countWords: false }
h1: { score: 1, countWords: true }
h2: { score: 1, countWords: true }
h3: { score: 1, countWords: true }
h4: { score: 1, countWords: true }
h5: { score: 1, countWords: true }
h6: { score: 1, countWords: true }
a: { score: 5, countWords: true }
li: { score: 0.5, countWords: true }
ul: { score: 1, countWords: true }
td: { score: 0, countWords: true }
hr: { score: 0, countWords: true }
pre: { score: 0, countWords: false }
ol: { score: 1, countWords: true }
wordValue: 0.1
- role: ["PULL_SPECIFICATION"]
multiplier: 0
rewards:
html:
br: { score: 0, countWords: true }
code: { score: 5, countWords: false }
p: { score: 0, countWords: true }
em: { score: 0, countWords: true }
img: { score: 5, countWords: true }
strong: { score: 0, countWords: false }
blockquote: { score: 0, countWords: false }
h1: { score: 1, countWords: true }
h2: { score: 1, countWords: true }
h3: { score: 1, countWords: true }
h4: { score: 1, countWords: true }
h5: { score: 1, countWords: true }
h6: { score: 1, countWords: true }
a: { score: 5, countWords: true }
li: { score: 0.5, countWords: true }
ul: { score: 1, countWords: true }
td: { score: 0, countWords: true }
hr: { score: 0, countWords: true }
pre: { score: 0, countWords: false }
ol: { score: 1, countWords: true }
wordValue: 0
- role: ["PULL_AUTHOR"]
multiplier: 2
rewards:
html:
br: { score: 0, countWords: true }
code: { score: 5, countWords: false }
p: { score: 0, countWords: true }
em: { score: 0, countWords: true }
img: { score: 5, countWords: true }
strong: { score: 0, countWords: false }
blockquote: { score: 0, countWords: false }
h1: { score: 1, countWords: true }
h2: { score: 1, countWords: true }
h3: { score: 1, countWords: true }
h4: { score: 1, countWords: true }
h5: { score: 1, countWords: true }
h6: { score: 1, countWords: true }
a: { score: 5, countWords: true }
li: { score: 0.5, countWords: true }
ul: { score: 1, countWords: true }
td: { score: 0, countWords: true }
hr: { score: 0, countWords: true }
pre: { score: 0, countWords: false }
ol: { score: 1, countWords: true }
wordValue: 0.2
- role: ["PULL_ASSIGNEE"]
multiplier: 1
rewards:
html:
br: { score: 0, countWords: true }
code: { score: 5, countWords: false }
p: { score: 0, countWords: true }
em: { score: 0, countWords: true }
img: { score: 5, countWords: true }
strong: { score: 0, countWords: false }
blockquote: { score: 0, countWords: false }
h1: { score: 1, countWords: true }
h2: { score: 1, countWords: true }
h3: { score: 1, countWords: true }
h4: { score: 1, countWords: true }
h5: { score: 1, countWords: true }
h6: { score: 1, countWords: true }
a: { score: 5, countWords: true }
li: { score: 0.5, countWords: true }
ul: { score: 1, countWords: true }
td: { score: 0, countWords: true }
hr: { score: 0, countWords: true }
pre: { score: 0, countWords: false }
ol: { score: 1, countWords: true }
wordValue: 0.1
- role: ["PULL_COLLABORATOR"]
multiplier: 1
rewards:
html:
br: { score: 0, countWords: true }
code: { score: 5, countWords: false }
p: { score: 0, countWords: true }
em: { score: 0, countWords: true }
img: { score: 5, countWords: true }
strong: { score: 0, countWords: false }
blockquote: { score: 0, countWords: false }
h1: { score: 1, countWords: true }
h2: { score: 1, countWords: true }
h3: { score: 1, countWords: true }
h4: { score: 1, countWords: true }
h5: { score: 1, countWords: true }
h6: { score: 1, countWords: true }
a: { score: 5, countWords: true }
li: { score: 0.5, countWords: true }
ul: { score: 1, countWords: true }
td: { score: 0, countWords: true }
hr: { score: 0, countWords: true }
pre: { score: 0, countWords: false }
ol: { score: 1, countWords: true }
wordValue: 0.1
- role: ["PULL_CONTRIBUTOR"]
multiplier: 0.25
rewards:
html:
br: { score: 0, countWords: true }
code: { score: 5, countWords: false }
p: { score: 0, countWords: true }
em: { score: 0, countWords: true }
img: { score: 5, countWords: true }
strong: { score: 0, countWords: false }
blockquote: { score: 0, countWords: false }
h1: { score: 1, countWords: true }
h2: { score: 1, countWords: true }
h3: { score: 1, countWords: true }
h4: { score: 1, countWords: true }
h5: { score: 1, countWords: true }
h6: { score: 1, countWords: true }
a: { score: 5, countWords: true }
li: { score: 0.5, countWords: true }
ul: { score: 1, countWords: true }
td: { score: 0, countWords: true }
hr: { score: 0, countWords: true }
pre: { score: 0, countWords: false }
ol: { score: 1, countWords: true }
wordValue: 0.1
permitGeneration: {}
githubComment:
post: true
debug: false
Partner private key (evmPrivateEncrypted
config param in conversation-rewards
plugin) supports 2 formats:
PRIVATE_KEY:GITHUB_OWNER_ID
PRIVATE_KEY:GITHUB_OWNER_ID:GITHUB_REPOSITORY_ID
Here GITHUB_OWNER_ID
can be:
- GitHub organization id (if ubiquity-os is used within an organization)
- GitHub user id (if ubiquity-os is simply installed in a user's repository)
Format PRIVATE_KEY:GITHUB_OWNER_ID
restricts in which particular organization (or user related repositories)
this private key can be used. It can be set either in the organization wide config either in the repository wide one.
Format PRIVATE_KEY:GITHUB_OWNER_ID:GITHUB_REPOSITORY_ID
restricts organization (or user related repositories) and a particular repository where private key is allowed to be used.
How to encrypt for you local organization for testing purposes:
- Get your organization (or user) id
curl -H "Accept: application/json" -H "Authorization: token GITHUB_PAT_TOKEN" https://api.github.com/orgs/ubiquity
- Open https://keygen.ubq.fi/
- Click "Generate" to create a new
x25519_PRIVATE_KEY
(which will be used in theconversation-rewards
plugin to decrypt encrypted wallet private key) - Input a string in the format
PRIVATE_KEY:GITHUB_OWNER_ID
in thePLAIN_TEXT
UI text input where:
PRIVATE_KEY
: your ethereum wallet private key without the0x
prefixGITHUB_OWNER_ID
: your github organization id or user id (which you got from step 1)
- Click "Encrypt" to get an encrypted value in the
CIPHER_TEXT
field - Set the encrypted text (from step 5) in the
evmPrivateEncrypted
config parameter - Set
X25519_PRIVATE_KEY
environment variable in github secrets of your forked instance of theconversation-rewards
plugin