Proposals/canton-foundation/canton-dev-fund/#439GitHub ↗

Declined#439 Pull request

Proposal: CCpedia - Canton Knowledge Infrastructure

Opened 09-06-2026 by 0xFearless-1, last activity 20d ago

Incoming09-06-2026
Declineddate not recorded

Funding request: 325,000 CC guaranteed + up to 605,000 CC adoption-based (tiered) = maximum 930,000 CC
Champion named in file: Digital Asset
Discussion: 6 messages; 2 people

Proposal file: /proposals/2026-06-UnityNodes-ccpedia.md

Summary

ccpedia is Canton's live knowledge infrastructure: 50,000+ records from 15 sources, 37,000+ RAG chunks, and 22 MCP tools already running in production. Phase 2 expands to 60 tools across 17 sources, open-sources the infrastructure under Apache 2.0, and gates 65% of funding on verified adoption by independent Canton teams (tiered: 3 / 5 / 7+ teams).

Checklist

Proposal file added under /proposals/
Milestones and funding amounts defined
Acceptance criteria included
Alignment with Canton priorities described

Milestones 3

Open-Source Release of Live Production InfrastructureUpon grant approval
Phase 2 Technical DeliveryMonth 3
Ecosystem Adoption6-month window after M2 acceptance

Discussion6 comments, 0 reviews, 0 inline comments

3 automated messages from bots hidden. Show bot messages

Showing 1 to 6 of 6, oldest first.

0xFearless-1commented09-06-2026 · 4mo agoGitHub ↗

CC @Andrew-Pohl @LimKianAn @Denend @nycnewman @mziolekda @monsieurleberre @hrischuk-da @paulbrauner-da @v9n

0xFearless-1commented11-06-2026 · 4mo agoGitHub ↗

What ccpedia is: a single searchable layer over the 15 places Canton info lives in today (forum, mailing lists, both GitHub orgs, docs, YouTube, dev fund, whitepapers, ecosystem, canton.foundation). The mailing list archives, the dev fund history, and the YouTube transcripts aren't indexed anywhere else. Who actually uses it: community devs onboarding to Canton, people writing about Canton (the Messari research report is one example, we just integrated it), Foundation DevRel who want to track what changes week to week. AI tooling is one consumer of the index, not the goal.

What's live right now at ccpedia.xyz (you can see all of this at /api/v1/sources): 15 sources, 33 MCP tools, around 39k RAG chunks, 124 CIPs, 4,147 forum topics, 2,885 doc pages, 880 mailing messages across 6 groups (cip-discuss, cip-vote, cip-announce, globalSyncForum, grants-discuss, validator-announce), 443 dev fund items, 165 YouTube videos with 51 transcribed (the other 114 are still being processed because YouTube rate-limits cloud IPs), 113 GitHub releases, 184 ecosystem projects, 41 canton.foundation pages, plus the 150 entries from Jatin's Build-on-Canton-MCP Curated KB that we integrated after he asked us to, with attribution preserved.

In the past week the MCP server grew from 22 to 33 tools. The new ones are list_videos, semantic_search, get_cip_votes, list_mailing_threads, get_mailing_thread, get_foundation_info, plus 5 tools wired to Jatin's KB. We also added the Daml YouTube channel (83 historical videos going back to 2019), the Messari research report, the cip-vote and cip-announce mailing lists, and canton.foundation as a doc source. Transcripts now come straight from YouTube's internal API, no yt-dlp and no cookies.

What the funding covers: keeping the sync running across all sources, embeddings and LLM inference, hosting, and the ongoing work of adding new sources as they show up, new MCP tools, search improvements, basically keeping ccpedia aligned with how Canton itself keeps changing. The full open-source release under Apache 2.0 is the M1 deliverable once the grant is approved.

CC @waynecollier-da @hythloda @stas-sbi @tkatrichenko @isegall-da @Andrew-Pohl @LimKianAn @Denend @nycnewman @mziolekda @monsieurleberre @hrischuk-da @paulbrauner-da. If anyone wants a walk-through of ccpedia.xyz, or has a specific section you want more detail on, just let me know. Happy to point at concrete examples so you can form a read.

0xFearless-1commented17-06-2026 · 4mo agoGitHub ↗

Milestone 1 - scope update (live infrastructure has grown since submission)

Already live in production at ccpedia.xyz - the code is closed-source today; the MCP service runs in production now. Every number below is independently verifiable at https://ccpedia.xyz/api/v1/sources:

33 MCP tools live in production (up from 22 at submission)
15 live data sources, syncing continuously
14,630 indexed records + 34,452 RAG chunks (~49k total knowledge units)
166 Canton YouTube videos indexed and searchable

The 33 production tools M1:

Search & retrieval: search, semantic_search, get_doc, get_api_reference
CIPs: get_cip, list_cips, get_cip_history, get_cip_attachments, get_cip_mentions, get_cip_votes
Forum / mailing / GitHub discussions: get_discussion, list_mailing_threads, get_mailing_thread, list_github_discussions, get_github_discussion
Whitepapers: get_whitepaper, list_whitepapers
Network & market: get_network_stats, get_network_state, get_token_market
Ecosystem & repos: list_ecosystem_projects, list_repos, get_foundation_info
Proposals: get_proposals
Videos: get_video, list_videos
Changes / radar / trending: get_recent_changes, get_radar, get_trending
Developer guidance: get_started_guide, get_faq, check_deprecation, compare_to_evm

Jatinp26membercommented23-06-2026 · 4mo agoGitHub ↗

thoughts @0xFearless-1

~140K USD is not properly broken down for only a MCP tool OR Public good API, Need proper breakdown on it.
Champion: Digital Asset is added, do you have consent of anyone at DA before adding them as champion??? If yes please mention them too.
Overall, there are many open source and free public good MCPs and more claude skills etc available. How'd this be diff to them? and if tomorrow they added same list of sources of context into their MCPs, what'd be the differentor here?

The tool is of value yes, but as a tool whose alternatives are available for free and grant to make it although its running already, that'd be conflicting for me to understand.

Suggestion: Instead of grant to make an already running MCP service, better approach is to explore options of only maintain it for public good out there.

0xFearless-1commented26-09-2026 · 1mo agoGitHub ↗

Thanks for the Dev Fund 2.0 notice. This PR is still the June version. We're keeping it as is until a champion is on board, so the rewrite can reflect their feedback. Here is where we stand.

RFP category: none, we see this as an individual initiative outside the roadmap. The closest is #15 (Training and Documentation), but ccpedia is a running service rather than authored material.

Why it's needed: Canton knowledge lives in many places (forum, mailing lists, GitHub, docs, video). ccpedia gives developers and their AI assistants one searchable, cited view of it. It's live at ccpedia.xyz: 90 MCP tools and 19 sources today.

Champion: we're still looking for one. The "Champion: Digital Asset" line and the label that followed it aren't confirmed on DA's side, so please treat us as still looking. @Jatinp26 , we haven't forgotten your June questions and will answer them in the rewrite.

0xFearless-1commented06-10-2026 · 20d agoGitHub ↗

Hi @Jatinp26, thank you for the detailed review in June, and for your patience. You asked fair questions and I owe you straight answers before I close this PR.

Budget breakdown: you were right. The file splits funding by milestone (OSS release, Phase 2 build, adoption tiers) but never by cost: engineering time, hosting, embeddings, inference. M1 asked for payment upfront for infrastructure that was already running, which is the weak point you identified.

Champion: the "Champion: Digital Asset" line was never confirmed with anyone at Digital Asset, and I should not have written it. The champion-confirmed label followed from that line and was not accurate. Apologies for the noise this caused the committee.

Differentiation: honestly, the list of sources is not a moat. If other MCPs add the same sources tomorrow, that is a good outcome for developers. What ccpedia offers is maintained, cited access to material that is hard to find elsewhere (mailing list archives, dev fund history, video transcripts, CIPs and votes), kept in sync continuously. That is operations, not something that needs exclusivity.

Your suggestion: you proposed keeping it running as a public good instead of funding a build of something already live. We agree. Unity Nodes will keep operating ccpedia.xyz as a free service on its own budget, and the Apache 2.0 repository promised in M1 is already public at github.com/UnityNodes/ccpedia. Today it covers 19 sources and 90 MCP tools (ccpedia.xyz/api/v1/sources).

So I am withdrawing this proposal and closing the PR. It does not map to a Dev Fund 2.0 RFP category as written, and a rewrite without a champion would not help the committee. If a champion from the ecosystem wants to sponsor a scoped proposal later, we will open a fresh PR in the new format.

Thank you to everyone on the committee for the time you spent on this.

Proposal file proposals/2026-06-UnityNodes-ccpedia.md, 18 KB

Development Fund Proposal

Author: Unity Nodes Status: Submitted Created: 2026-06-04 Label: daml-tooling Champion: Digital Asset

Abstract

ccpedia (ccpedia.xyz) is Canton's knowledge infrastructure: a live, unified corpus of 50,000+ records from 15 sources, 37,000+ RAG chunks, and 22 MCP tools already running in production (record counts independently verifiable at ccpedia.xyz/api/v1/sources). This proposal funds Phase 2: expanding to 60 MCP tools across 17 data sources, including ten cross-source synthesis tools that are only possible because of ccpedia's unique unified corpus. By month 9, ccpedia is open-sourced under Apache 2.0, expanded to 60 tools across 17 sources, and in verified active use by independent Canton teams.

Specification

1. Objective

Canton developers lack a unified, always-current knowledge layer. Finding answers requires searching docs, forum posts, GitHub issues, CIPs, proposals, and videos separately, often returning contradictory or outdated results. Without a unified solution, AI coding assistants (Claude, Cursor, Windsurf) remain blind to Canton's living ecosystem: no single MCP-accessible layer synthesizes all sources into one coherent whole.

ccpedia solves this today: a single MCP server aggregating, syncing, and semantically searching 15 Canton knowledge sources in production. Phase 2 expands the existing 22 live tools to 60 tools across 17 sources, adding cross-source synthesis capabilities that are fundamentally impossible without ccpedia's unified corpus.

Single objective: Expand ccpedia from a working tool to the canonical Canton developer knowledge infrastructure, open-sourced under Apache 2.0, with verified external adoption.

2. Implementation Mechanics

The work is structured across three milestones, each building on the previous.

M1: Open-Source Release of Live Production Infrastructure (upon grant approval)

M1 releases live production infrastructure as a Canton ecosystem public good under Apache 2.0:

22 MCP tools in production, 15 live data sources, sync infrastructure running continuously
Open-source release at github.com/UnityNodes/ccpedia (Apache 2.0): MCP server, all tools, database schema, sync pipelines, source connectors
45 Canton YouTube videos indexed and searchable via the existing search and get_video tools
All 22 existing tools documented with usage examples

M2: Phase 2 Technical Delivery (Month 3)

38 new MCP tools total: 22 per-source, 10 synthesis, and 6 new-source across 2 new data sources.

22 new per-source MCP tools targeting HIGH/MEDIUM developer demand, including:

diagnose_error(error_text): searches 4,100+ forum topics and 3,100+ GitHub issues for matching errors and community workarounds
get_breaking_changes(from_sdk, to_sdk): what breaks when upgrading between SDK versions (440+ versions indexed)
get_sdk_changelog(version): detailed per-version changelog
search_community(query): semantic search across forum, mailing list, blog, and discussions in one call
search_github_issues(query): full-text search across 3,100+ GitHub issues from 15+ repos
search_talks(query): semantic search across 45 YouTube videos with automated transcripts
find_known_issues(description), get_faq(topic), find_cip_for_feature(feature), get_cip_implementation_status(cip_number), get_latest_release(package), search_release_notes(feature), find_security_patches(), get_npm_packages(query), find_similar_projects(description), search_mailing_list(query), find_maintainer_guidance(topic), get_issue_status(issue_number), find_code_examples(topic), get_funding_landscape(category), get_proposal_milestones(id), get_ecosystem_gaps()

10 cross-source synthesis tools requiring ccpedia's unified corpus (no individual source supports these):

detect_builder_overlap(idea): identifies who is building something similar across 410+ proposals and 180+ ecosystem projects; manual research takes 3+ hours, this tool takes 10 seconds
proposal_success_predictor(draft_text): heuristic scoring against funded vs. unfunded attributes of all 410+ historical proposals; the manual cross-proposal analysis that this tool will automate was itself applied when writing this proposal
full_context(topic): single query across docs, forum, proposals, CIPs, blog, and GitHub simultaneously
community_consensus(topic): synthesizes what the Canton community actually believes, not just the last forum reply
find_expert(topic): algorithmic expert ranking by forum reputation, GitHub contributions, and proposal authorship
detect_drift(topic): finds contradictions between docs, forum, and GitHub (for example, docs say one thing, forum says another, and explains why)
learning_path(goal): structured onboarding path from whitepaper through docs, forum, proposals, and videos
outdated_guidance_detector(query): flags stale 2023 advice and surfaces the current equivalent
ecosystem_dependency_graph(topic): maps 180+ projects to SDK versions, repos, and proposals
find_collaboration_opportunities(project): surfaces projects with complementary capabilities

2 new live data sources:

Featured Apps via Scan API (16th source): list_featured_apps, get_app_metrics, find_apps_by_pattern
SV Operations Calendar (17th source): Canton Foundation ICS feed (sv-cal.canton.foundation/schedule.ics, auto-updates every 12 hours) with get_sv_schedule(environment), get_upcoming_operations(days), get_recent_deployments()

External code review and penetration test of the ccpedia MCP server, a read-only HTTP API/RAG service. Threat model scope: prompt injection protection, rate limiting, data leakage, query sanitization. This is not a smart contract audit (ccpedia holds no keys, executes no contracts, manages no funds). Public docs at ccpedia.xyz/docs, developer onboarding guide for Claude/Cursor/Windsurf, 20 curated Canton design patterns with code examples.

M3: Ecosystem Adoption (6-month window after M2 acceptance)

Verified adoption of ccpedia by independent Canton teams. Unity Nodes runs an active promotion campaign (Discord, Slack, Forum, X) and direct outreach to Canton builder teams. Gate: at least 3 independent Canton teams publicly document active use of ccpedia via Canton forum post, GitHub issue/PR, published case study, or written statement shared with the committee.

Technical stack: SQLite with sqlite-vec for RAG (vector similarity search chosen for zero-dependency local vector search without an external managed service), Next.js API layer, MCP protocol for tool exposure, incremental sync pipelines (5 min / 15 min / 6 hr / daily per source).

3. Architectural Alignment

MCP standard: MCP tooling for Canton is an active area of committee engagement. ccpedia focuses on knowledge access; existing MCP implementations in the ecosystem focus on action-execution or DAML code analysis. These are complementary layers with no functional overlap.
Apache 2.0 open-source: Consistent with the open-source direction of the Canton Dev Fund.
CIP alignment: Makes 120+ CIPs semantically searchable with cross-references, implementation status, and related proposals. Directly supports developers implementing CIP-compliant applications.
Extension, not replacement: Phase 2 extends the live 22-tool system. No existing Canton tool or service is replaced or duplicated.
Canton Wiki differentiation: ccpedia is a live, semantically searchable corpus with 22 MCP tools, not a static wiki or registry.

4. Backward Compatibility

No backward compatibility impact. ccpedia is a read-only service layer that queries existing Canton data sources (forum, GitHub, docs, CIPs) without modifying them. The MCP server is purely additive infrastructure. The existing 22 tools remain unchanged; Phase 2 adds tools alongside them.

Milestones and Deliverables

Milestone 1: Open-Source Release of Live Production Infrastructure

Estimated Delivery: Upon grant approval
Payment: 185,000 CC, paid at grant approval
Focus: Open-source release of live production infrastructure as a Canton ecosystem public good
Deliverables / Value Metrics:
- Phase 1 work already delivered: 22 MCP tools in production, 15 live data sources, sync infrastructure running continuously
- Dedicated open-source repository (MCP server, sync pipelines, schema) published under Apache 2.0 at github.com/UnityNodes/ccpedia
- 45 Canton YouTube videos indexed and searchable via the existing search and get_video tools
- All 22 existing tools documented with at least 1 usage example each

Milestone 2: Phase 2 Technical Delivery

Estimated Delivery: Month 3
Payment: 140,000 CC upon committee acceptance
Focus: Full Phase 2 technical build: 38 new tools (22 per-source, 10 synthesis, 6 new-source), 2 new data sources, code review, docs
Deliverables / Value Metrics:
- 22 new per-source MCP tools operational and documented
- 10 cross-source synthesis tools live and demonstrably functional via live demo
- 6 new-source tools live (Featured Apps: 3, SV Calendar: 3)
- 2 new data sources live and syncing (Featured Apps via Scan API, SV Operations Calendar)
- External code review completed, findings remediated, full report publicly accessible
- ccpedia.xyz/docs live with tool reference and integration guides for Claude / Cursor / Windsurf
- 20 Canton design patterns published at ccpedia.xyz/docs

Milestone 3: Ecosystem Adoption

Estimated Delivery: 6-month window after M2 acceptance
Payment: Tiered based on verified adoption: 300,000 CC (3 teams) / 450,000 CC (5 teams) / 605,000 CC (7+ teams)
Focus: Verified adoption of ccpedia by independent Canton teams
Deliverables / Value Metrics:
- Active promotion campaign run by Unity Nodes (Forum, X, Discord, Slack, direct outreach to builder teams)
- Independent Canton teams (not affiliated with Unity Nodes) publicly document active use of ccpedia in their development workflow
- Evidence accepted: Canton forum post, GitHub issue/PR, published case study, or written statement shared with the committee
Gate (tiered):
- 3 independent Canton teams publicly document active use → 300,000 CC
- 5 independent Canton teams publicly document active use → 450,000 CC
- 7+ independent Canton teams publicly document active use → 605,000 CC

Acceptance Criteria

The Tech & Ops Committee will evaluate completion based on:

Demonstrated Canton developer adoption (documented external users, public repos, case studies) rather than CI/CD test results
Live data sources syncing continuously at stated cadences with verifiable record freshness
Open-source repository publicly accessible under Apache 2.0 with working setup instructions

The acceptance criteria is to be based on value to the ecosystem and not delivery of an artifact. M1 delivers the OSS release of already-running infrastructure. M2 delivers the full Phase 2 technical scope. M3 demonstrates that ccpedia is genuinely used by independent Canton teams.

M1 gate: Open-source repository publicly accessible at github.com/UnityNodes/ccpedia under Apache 2.0; Phase 1 tools documented

M2 gate: 38 new tools operational (22 per-source, 10 synthesis, 6 new-source for Featured Apps/SV Calendar); 2 new data sources syncing; code review report public; ccpedia.xyz/docs live

M3 gate (tiered): Independent Canton teams publicly document active use via forum post, GitHub issue/PR, published case study, or written statement shared with the committee: 3 teams → 300,000 CC / 5 teams → 450,000 CC / 7+ teams → 605,000 CC

Funding

Total Funding Request: 325,000 CC guaranteed + up to 605,000 CC adoption-based (tiered) = maximum 930,000 CC

Milestone	CC	When	Basis
M1: OSS Release of Live Infrastructure	185,000	Upon grant approval	Live production infrastructure published as Apache 2.0 public good
M2: Phase 2 Technical Delivery	140,000	Month 3, upon committee acceptance	Delivery of 38 new tools, 2 sources, code review, docs
M3: Ecosystem Adoption	300,000–605,000	6-month window after M2	Tiered: 3 teams (300k) / 5 teams (450k) / 7+ teams (605k)
Maximum total	930,000		20% OSS release / 15% technical / up to 65% adoption

Co-Marketing

Upon release of each milestone, Unity Nodes will collaborate with the Foundation on:

Announcement coordination for the Apache 2.0 open-source release (M1)
Technical blog post documenting ccpedia architecture, MCP integration patterns, and RAG pipeline design
Developer guide "Using ccpedia with Claude / Cursor / Windsurf" published at ccpedia.xyz/docs and submitted to Canton developer resources
Case study documenting how manual cross-proposal analysis, the research approach that proposal_success_predictor will automate, was applied when writing this proposal; published at ccpedia.xyz/docs after M2 delivery
Upon M2 completion, Unity Nodes will build and open-source a Discord/Slack bot that exposes ccpedia MCP tools in community channels, including a daily Canton digest (new CIPs, proposals, SDK releases) and interactive Q&A via ccpedia tools. The bot will be available for self-hosting by the Foundation or any community member. If requested, Unity Nodes can assist with deployment in official Canton community spaces.

Motivation

Canton developers waste dozens of hours monthly solving the same problems repeatedly: "What broke in this SDK upgrade?" "Who is already building this?" "What does the community actually think about X?" Each answer requires searching 4-6 separate sources, reconciling contradictions, and hoping the answer is not in a 2023 forum post that references a deprecated API.

ccpedia already solves this for the 15 sources it covers today. Phase 2 extends the coverage and adds synthesis: the ability to answer questions that require reasoning across the entire corpus simultaneously.

Ecosystem reach: Canton developers using AI coding assistants (Claude, Cursor, Windsurf); grant proposal teams doing due-diligence on existing work; new developers needing a structured onboarding path; grant committee members verifying overlap between proposals.

Proof of work before asking for funding: 50,000+ live records from 15 sources, 37,000+ RAG chunks, 22 MCP tools in production, all sustained by Unity Nodes without a grant. Phase 2 is not a promise; it is an extension of something already running.

Rationale

Why ccpedia rather than extending an existing tool? No existing Canton tool provides unified semantic search across multiple live sources with MCP exposure. ccpedia extends what exists without replacing anything.

Why these 38 new tools specifically? The 22 new per-source tools (M2) were selected by mapping which Canton developer questions take the most time today: error diagnosis, SDK upgrade planning, and community consensus queries. The 10 synthesis tools (M2) were selected by identifying which queries require the entire corpus simultaneously (cross-source synthesis is impossible per-source). The 6 new-source tools for Featured Apps and SV Operations Calendar (M2) were selected based on confirmed public availability and developer demand for live dApp metrics and network operations scheduling. Additional tool candidates are deferred to a subsequent grant based on real-world Phase 2 adoption data.

Why code review and not a full security audit (M2)? ccpedia is a read-only HTTP API and RAG search service. It holds no keys, executes no smart contracts, and manages no user funds. The threat model is prompt injection, data leakage, and query sanitization, which is appropriate for a code review and penetration test scope. A full smart contract audit would be disproportionate to the actual risk surface and would consume budget better spent on adoption-driving productization.

Why Apache 2.0 open-source in M1, and why is M1 paid upon grant approval? Unity Nodes invested significant time and resources building Phase 1 (22 tools, 15 live data sources, production infrastructure) without prior grant funding. M1 funds the OSS release of infrastructure already running in Canton, documenting and licensing it for public reuse, converting this investment into a permanent public Canton ecosystem asset. The entire Canton community gains the ability to verify, fork, and self-host ccpedia independently of Unity Nodes. The software is open-source under Apache 2.0; the managed data service (indexed corpus, live sync pipelines, and RAG layer) remains operated by Unity Nodes. The repository will be published at github.com/UnityNodes/ccpedia.

Why this funding amount? The 930,000 CC maximum reflects three distinct scopes of work: M1 (185,000 CC) funds the OSS release of Phase 1 infrastructure (22 tools, 15 sources, production sync pipelines) as a Canton ecosystem public good; M2 (140,000 CC) funds 38 new tools, 2 new data sources, external code review, and public documentation over 3 months of active development; M3 (up to 605,000 CC) is tiered based on verified external adoption: 300,000 CC for 3 teams, 450,000 CC for 5 teams, 605,000 CC for 7+ teams — aligning ccpedia's financial upside directly with ecosystem outcome. The structure is 20% OSS release / 15% technical delivery / up to 65% adoption-based: the adoption tranche scales with the number of independent Canton teams that publicly document active use.

Why Apache 2.0 specifically? Apache 2.0 includes a patent grant clause that protects users and contributors from patent litigation. It matches the license of upstream Canton dependencies (canton, daml, splice are all Apache 2.0), eliminating license compatibility friction for downstream integrators.

Why a performance-based M3? Unity Nodes can demonstrate that ccpedia tools work. Only the Canton developer community can demonstrate that ccpedia tools are actually used. The M3 structure (verified external teams with publicly documented use) aligns ccpedia's financial upside with the outcome that matters to the ecosystem.

← Back to Proposals Open on GitHub ↗