← Home
NB
Nervos Brain
01 / 12

Community Testing & Delivery Report

Nervos Brain

Spark Program delivery summary: a deployable Nervos / CKB community knowledge system prototype with retrieval, citations, reply context, and rating feedback.

Spark Delivery ReadyTesting ClosedJune 2026
60Real Answers
27Ratings
8Real Testers
40Supplemental Runs
Spark Delivery Pillars
UsableValidated with real Telegram community questions
Evidence-BasedGrounded in docs, Talk archives, and GitHub code retrieval
MeasurableRatings, failure cases, and regression cases collected
Delivery Evidence
Real Group Test60 answers recorded
Rating Feedback27 ratings collected
Supplemental Runs40 fixed validation runs
Knowledge System
Docs
Talk
GitHub
Brain
Intelligent RAG Q&AQdrantArchive DBTelegram / Discord
24Evidence-backed Answers
13Follow-up Cases
Product Delivery Dashboard

Executive Summary

A deployable, answerable, feedback-ready community knowledge system prototype is complete.

02 / 12
Delivery Status
Ready

The current version is ready for Spark community testing delivery. The DAO phase can focus on speed, coverage, and verifiability.

Telegram LiveFeedback LiveDocs Ready
12Report Slides
8Docs Modules
Quality Result
3.67 / 5

Average across all real-group ratings; aggregated by answer: 3.75 / 5.

Coverage
~16 Perspectives

8 real users + 8 supplemental test personas.

8Real
8Personas
Feedback Loop
Ask@ mention or reply
AnswerRetrieval + citations
Ratings1-5 satisfaction

Rating buttons and failure-case logging were validated in the real group.

27Ratings
24Rated Answers
Validated Scope
CKB BasicsFiberCCCWallet SafetyTalkSpore/DOB
6Validated Areas
2DAO Deepening Areas
Score Structure

4-5 ratings account for 63.0%; lower scores mainly mark data gaps, boundary cases, and noisy test messages.

17High-score Answers
105-star Ratings
Executive Summary

System Capabilities

The core loop is connected, from chat entry points and retrieval stores to feedback.

03 / 12
Runtime Flow
Telegram / DiscordCommunity Entry
Graph EngineRouting + Planning
RetrieverDocs / Talk / Code
Answer GenerationAnswer
Citation QualitySources
SatisfactionFeedback
Reply ContextLanguage FollowingLong-message SplittingProgress UpdatesFormat Repair
Input@ mentions or replies from community chats
Evidence AccessDocs, Talk archives, and GitHub examples
OutputAnswers, citations, and rating buttons
Knowledge Base Map
DocsConcepts + Tutorials
Nervos TalkCommunity History
GitHub CodeExamples + Repos
3Collections
LFSArchive DB
Community Entry
2 Platforms
Telegram
Discord

Current entry points are working; the long-term positioning is a knowledge system.

Data Model
3 Collections
Docs
Forum
Code

Docs, forum content, and GitHub code form the retrieval foundation.

Operations
Docs Ready
Deployment
Config
Data
Run
Troubleshoot

Deployment, configuration, data, runtime, and troubleshooting docs are complete.

Update Capability
Incremental Updates
Talk
GitHub

Talk and GitHub refresh paths are ready.

Capability Overview

Test Coverage

Real group testing + supplemental runs cover about 16 testing perspectives.

04 / 12
Tester Perspective Coverage
~16Perspectives
8Real Users
8Personas
Answers by Date

The official test group recorded 60 answers; June 22 added 4 answers from 2 testers that day.

60Real Answers
13Follow-up Cases
Coverage Mix
Real Testers8
Supplemental Personas8
Supplemental Runs40
Tester Types
CKB DevelopersCCC / TSFiberSpore / DOBWeb3 DevelopersBeginnersWallet UsersCommunity Ops
Question Types
Beginner QuestionsTutorial QuestionsCode / SDKTalk / CommunityReply Follow-upsStrategy Discussion
8 Real Testers + 8 Supplemental Personas

Quality Result

Both real ratings and supplemental runs show that 4-5 scores dominate.

05 / 12
Real Avg.
3.67 / 5

27 ratings.

By Answer
3.75 / 5

24 rated answers.

High-score Share
63.0%

4-5 score share.

Supplemental Avg.
3.83 / 5

40 fixed runs.

Real-group Ratings 3.67 / 5

5
10
4
7
3
5
2
1
1
4

Supplemental Run Ratings 3.83 / 5

5
5
4
25
3
8
2
2
1
0
Positive Signal
17 / 27

17 of 27 real ratings are 4-5, or 63.0%.

Low-score Reading
Data GapsBoundary CasesTest Noise

Lower scores mainly locate boundaries and data gaps; they do not mean the core Q&A is unusable.

Acceptance Judgment
Sufficient for Spark Delivery

Real testing and supplemental runs agree: a usable MVP has been established.

Quality Summary

Response Time

Direct answers are fast; complex RAG questions are the DAO-phase optimization focus.

06 / 12
Direct Answers
36 answers
Median 18.1s
P90 41.4s
Normal RAG
24 retrieval answers
Median 96.6s
P90 199.7s
Supplemental Runs
40 local runs
Median 17.6s
P90 118s
Fast Path WorksSimple questions respond quickly
RAG Path WorksSlower, but evidence-backed
Next LeverControl reflection and citation assembly
Edge-case Long Chain
One extreme case observed

The xUDT contract-review run took 1104.2s originally; retest was about 133.75s, while retrieval itself was about 0.8s. This looks more like an occasional long generation, reflection, self-check, and citation-assembly chain.

Original Run1104.2s
Retest133.75s
Retrieval Time~0.8s
Interpretation
Not a Delivery BlockerDAO-phase Optimization

Only one extreme long-chain case appeared across multiple testing days, and the retest dropped significantly.

Optimization Levers
Time BudgetBound post-reflectionCitation Parsing

The main lever is not retrieval speed, but bounding generation, reflection, self-check, and citation assembly.

DAO Goal
Predictable Latency

Move complex questions from “can answer” to “answers reliably within a controlled time budget.”

Response Time Breakdown

Capability Validation

Core capabilities are validated; a few vertical topics move into the DAO-phase backlog.

07 / 12
Capability Heatmap
Area
Accuracy
Citation Quality
Follow-up
DAO Next
CKB Basics
Strong
Optional
Stable
Citations
CCC / TypeScript
Useful
Examples
Stable
Runnable Code
Fiber
Strong
Supported
Stable
Fresh Data
Spore / DOB
Usable
Needs More
Usable
Coverage
Talk / Community
Useful
Supported
Usable
Metadata
Validation Signals
Follow-up Cases13
Evidence-backed Answers24
Split Messages1
Language FollowingFixed
8Capabilities
DAODeepening Backlog
Sufficient for Spark Delivery
CKB OnboardingTechnical ExplanationFiber / CCC PathsWallet SafetyReply ContextFeedback CollectionSpore/DOB DepthComplete Latest Lists
Capability Matrix

Representative Strong Cases

High-score cases cover technical explanation, deployment paths, safety guidance, and product discussion.

08 / 12
C
5 / 5 · Technical

CKB Cell & Scripts

Explains cells, lock scripts, transaction flow, and dead cells at the coding level, with stable multi-turn context.

Accurate ConceptsStable Follow-upsClear Structure
F
4 / 5 · Deployment

Fiber Node Agent Deployment

Covers node, wallet, channels, RPC / SDK control, and key-isolation guidance.

Complete PathSafety BoundaryActionable
T
4 / 5 · Dev Tooling

CCC TypeScript Transfer

Organizes the answer around project setup, transaction construction, input completion, fee completion, and sending.

StepwiseDeveloper-friendlyReusable
S
5 / 5 · Safety

Wallet Safety

Refuses private keys and seed phrases clearly, while giving safer alternatives for scam checks.

Risk RecognitionCorrect RefusalSafer Alternatives
L
5 / 5 · Limit Disclosure

Data-limit Disclosure

Clearly states it cannot browse the latest web pages and recommends checking official primary sources.

Transparent LimitsNo FabricationSource Awareness
P
4-5 / 5 · Strategy

Product Discussion

Provides structured analysis for CKB growth, prediction markets, cold start, and risk boundaries.

Complete LogicRisk-awareDiscussion-ready
Representative High-quality Cases

Test -> Fix -> Delivery

Internal testing drove multiple productization fixes; the delivery build is more stable than the recruitment build.

09 / 12
Fix Dashboard
Language FollowingEnglish questions are no longer overridden by account language.Fixed
Progress-update LanguageProcessing updates follow the question language.Fixed
Reference HeadingReference-section headings adapt to Chinese / English.Fixed
Reply ContextAnswers prioritize the replied-to message.Validated
Citation QualityCitation format and mismatch issues are reduced.Improved
Retrieval RecallRecall for project names, filenames, and repos is improved.Improved
Speed TuningReduces unnecessary multi-round reflection.Ongoing
Reporting MaterialsTest report, speaker notes, and delivery materials are finalized.Ready
Iteration Impact
Language UXStable
Reply ContextFixed
Reference HeadingLocalized
Recall DepthImproved
7Fix Areas
1Ongoing Speed Item
Before Fixes
Testing Exposed Boundaries

Language, citations, reply context, and recall boundaries surfaced in real group chats.

Language MismatchObserved
Follow-up DriftObserved
Weak CitationsObserved
After Fixes
Delivery Build Hardened

Core visible issues are fixed; remaining speed work moves to the DAO backlog.

LanguageFixed
Reply ContextFixed
Reference HeadingLocalized
Product Result
More Reliable Usable Prototype

The delivery build is more stable than the initial recruitment-test version and better suited for acceptance.

Visible Fixes7
Remaining ItemSpeed
Delivery StatusReady
Product Iteration Record

Known Boundaries

Boundaries are now DAO-phase backlog items, not delivery blockers.

10 / 12
DAO Backlog Quadrant
Speed BudgetPriority
Vertical CoverageEnhance
Citation PrecisionEnhance
High-completeness TasksStructure
PriorityPrioritize speed and citation reliability
ScopeKnowledge-depth issues, not usability issues
OwnerDAO-phase product backlog
Scope Judgment
Spark Prototype Complete

The current version supports real group Q&A, basic technical explanation, data retrieval, reply follow-ups, and feedback collection. Boundary issues are suitable for continued DAO-phase work.

Non-blockingDAO RoadmapQuality Backlog
Prototype StatusAcceptable
Risk TypeEnhancement
Next OwnerDAO Phase
PrioritySpeed + Citations
Boundary Cases
Spark Full ListsStructured Data
Talk User Behavior DataMetadata
Spore / DOB Concept BoundariesVertical Docs
xUDT Contract-review EntrySpeed Budget
Short-confirmation IntentFollow-up Strategy
Latest Events / Web PagesSource Limits
DAO Enhancement Boundaries

DAO Roadmap

From Spark prototype to long-term community knowledge infrastructure.

11 / 12
Growth Path
Phase 1Stabilize & Speed Up
  • Time budget for complex chains
  • Bound post-reflection
  • Citation parsing mechanism
  • Automated regression tests
Phase 2Expand Coverage
  • Structured ecosystem project data
  • Talk metadata and governance data
  • More GitHub example entry points
  • Spore / DOB / xUDT vertical enhancement
Phase 3Community Knowledge Infrastructure
  • Community data contribution flow
  • Public evaluation and quality dashboard
  • Continuous incremental updates
  • DAO operations and feedback loop
North-star Goals
FasterComplex Q&A becomes more predictable, with budgets for long chains.
BroaderCover more ecosystem projects, Talk history, and GitHub examples.
More VerifiableCitations bind tightly to evidence and support public evaluation.
Community Co-buildingMove from one-time delivery to continuously updated knowledge infrastructure.
From Usable Prototype to Community Knowledge Infrastructure

Final Acceptance

The current version is suitable as the Nervos Brain community-testing delivery build.

12 / 12
Acceptance Summary
Spark Delivery Ready

Real group Q&A, follow-ups, and rating feedback are working; core Nervos / CKB knowledge Q&A has reached usable prototype quality; reports, test sets, deployment docs, and future enhancement directions are clear.

60Answers
27Ratings
8Real Testers
12Report Slides
Acceptance Checklist
Telegram group testing complete
About 16 testing perspectives covered
Feedback collection validated
Core Q&A capability validated
DAO enhancement backlog organized
Delivery Loop
01Community testing complete
02Feedback data collected
03Delivery materials ready
04DAO proposal handoff