Mirrored from filpgf.io — ProPGF Batch 3 (Karma program 1479, application 6a31814713ef2aeed7ce0d8f, status: pending). Contact details redacted; canonical application lives on filpgf.io. 1.1 Project Name Filecoin Machine-Readable Ecosystem Knowledge Base 1.2 Project Github https://github.com/neti-software/llm-assistant-for-code-repos 1.4 Team Lead/Point of Contact Damian Pudło, Blockchain and Compliance Expert, preferred channel: Filecoin Slack @Damian Pudło or via e-mail 1.5 Category [ "RFP 3 - AI infrastructure products on Filecoin" ] 1.6 Open Source Status Fully Open Source 2.1 Project Summary Project Description The goal of this project is to create and maintain a machine-readable Filecoin ecosystem knowledge layer that can be consumed by both developers and AI systems. Today, Filecoin technical knowledge is distributed across documentation sites, GitHub repositories, source code, specifications, issues, pull requests, release notes, community discussions, and operational know-how. This makes it difficult for developers, Storage Providers, application builders, and AI assistants to access accurate and up-to-date information. The problem is becoming more important as Filecoin grows across multiple products, pods, protocols, client-facing workflows, Storage Provider tooling, and AI-assisted development environments. Developers increasingly use AI coding assistants, RAG tools, and chat-based interfaces to understand technical systems. These tools are only useful if they can access current, structured, reliable, machine-readable knowledge. The proposed project will organize Filecoin ecosystem knowledge into a maintained, machine-readable knowledge base. It will combine source-code-aware indexing, documentation improvement, repository monitoring, release tracking, structured metadata, freshness checks, and AI-consumable content formats. The result will be a public-good knowledge infrastructure layer for the Filecoin ecosystem: useful for developers reading documentation, teams maintaining technical content, and AI systems that need reliable context to answer Filecoin-related questions or assist with development and operations. This proposal builds on the same underlying proof-of-concept as our Filecoin SP Operations Intelligence submission, but extends it in a different direction. The shared foundation is a Filecoin-focused AI knowledge system that indexes documentation, repositories, source code, and live project metadata. In Filecoin SP Operations Intelligence, this foundation is extended toward local environment awareness, Storage Provider diagnostics, operational health checks, and approved operational actions. In this proposal, the same foundation is extended toward ecosystem-level machine-readable knowledge: structured documentation, source-aware indexing, freshness tracking, AI-consumable outputs, and reusable knowledge infrastructure for developers, builders, documentation maintainers, and AI systems. Existing Proof-of-Concept This project builds on an existing proof-of-concept that already demonstrates the feasibility of the approach. The current system is a microservice-based application composed of three loosely coupled containers communicating over HTTP. The presentation layer is a Streamlit-based chat interface. It connects to the agent core through a gateway exposing an OpenAI-compatible chat/completions endpoint. The gateway is protected with token-based authentication and preserves session state across res …[truncated] 2.2 Who does this work support? [ "Application Builders", "Storage Providers", "Network Infrastructure", "Pods", "Application Users" ] 2.3 Total Funding Requested (USD) $92,000 2.4 Milestones & Budget [ { "title": "Milestone 1: Public Foundation and Knowledge Architecture", "description": "Estimated duration: Month 1\nDeliverables\nPublic GitHub repository.\nArchitecture documentation.\nSource list for initial Filecoin knowledge ingestion.\nKnowledge schema design.\nMetadata model.\nIngestion pipeline design.\nPublic roadmap.\nContribution guidelines.\nDocumentation explaining the proof-of-concept and planned production system.\nBudget Justification\nThis milestone funds the open-source foundation, project architecture, schema design, public documentation, and setup required for community review and future milestone validation.", "dueDate": "2026-08-30", "fundingRequested": "$16,000", "completionCriteria": "Acceptance Criteria\nPublic repository is available.\nArchitecture and roadmap are documented.\nInitial source list is published.\nKnowledge schema and metadata model are documented.\nContribution process is defined.\nProof-of-concept architecture is explained publicly." }, { "title": "Milestone 2: Filecoin Repository and Documentation Indexing", "description": "Estimated duration: Months 1–3\nDeliverables\nAutomated ingestion pipeline for selected Filecoin repositories.\nDocumentation scraping and Markdown conversion.\nOffline resumable indexing process.\nChunking strategy for code and technical documentation.\nEmbedding pipeline.\nVector index generation.\nSource metadata extraction.\nExclusion rules for forks, archived repositories, generated files, and low-value content.\nInitial indexed knowledge base covering key Filecoin repositories and documentation sources.\nBudget Justification\nThis milestone funds the core indexing system. The work is engineering-heavy because it must process code, documentation, metadata, and external sources while maintaining source traceability and freshness.", "dueDate": "2026-10-30", "fundingRequested": "$24,000", "completionCriteria": "Acceptance Criteria\nSelected Filecoin repositories are indexed.\nSelected documentation sources are indexed.\nEach indexed chunk includes source metadata.\nThe pipeline can resume after interruption.\nForks and archived repositories can be excluded.\nThe index can be rebuilt reproducibly.\nOutput can be consumed by the retrieval layer." }, { "title": "Milestone 3: Machine-Readable Knowledge Outputs and Retrieval API", "description": "Estimated duration: Months 3–4\nDeliverables\nAI-consumable knowledge exports.\nRetrieval API or query interface.\nOpenAI-compatible access pattern where applicable.\nSource-aware retrieval responses.\nVersion and freshness metadata in retrieval results.\nTopic and component tagging.\nExample integrations for AI assistants or RAG systems.\nPublic documentation for consuming the knowledge base.\nBudget Justification\nThis milestone turns the indexed knowledge into usable infrastructure for deve …[truncated] 3.1 Impact pathway **Output** → We will deliver an open-source, maintained, machine-readable Filecoin knowledge layer. The outputs include a public repository, source ingestion pipeline, indexed Filecoin repositories and documentation, structured metadata, source-aware retrieval, AI-consumable exports, freshness tracking, live project signals, evaluation datasets, documentation gap reports, and a public API or query interface for developers and AI systems. Outcome → Developers, Storage Providers, application builders, documentation maintainers, ecosystem teams, and AI assistants will be able to access more accurate, current, and source-backed Filecoin knowledge. Instead of relying on scattered documentation, outdated search results, or incomplete AI context, users will be able to query a maintained knowledge layer with source metadata, freshness signals, and structured outputs. This reduces onboarding friction, improves AI-assisted development quality, reduces repeated support questions, and helps builders find correct integration paths faster. **Impact** → Better machine-readable ecosystem knowledge improves Filecoin’s ability to scale adoption. Developers can integrate Filecoin services more confidently, Storage Providers can find operational knowledge more easily, and AI systems can provide more reliable Filecoin-specific answers. This supports the network’s 2026 objectives indirectly by improving the infrastructure that builders and operators rely on to create paid usage, support Storage Provider sustainability, and help customer-facing teams serve real users. The core change is that Filecoin knowledge moves from fragmented human-readable content into a structured, source-aware, machine-readable knowledge layer. That change makes the ecosystem easier to build on, easier to support, and easier for AI systems to understand. 3.2 Verification metrics This project does not directly create onchain transactions, storage deals, payments, or protocol-level state changes. Therefore, there are no direct onchain metrics that can be fully attributed to this work alone. However, the project will track measurable tool-level verification metrics that show whether the machine-readable knowledge layer is being built, maintained, used, and improved. These metrics can be externally reviewed through the public GitHub repository, public source lists, published metadata manifests, evaluation results, API documentation, GitHub issues, changelogs, and community feedback. **Verification Metrics** Number of Filecoin repositories indexed Target: index a prioritized set of key Filecoin repositories relevant to developers, Storage Providers, and ecosystem teams. Verification: public source list, repository metadata manifest, ingestion logs, and indexed-source documentation in the project repository. Number of Filecoin documentation sources indexed Target: index selected Filecoin documentation sources, product docs, operational guides, and relevant technical materials. Verification: public source list, documentation-source manifest, ingestion reports, and published knowledge base metadata. Percentage of indexed knowledge chunks with source metadata Target: at least 90% of indexed chunks include source URL or repository path, file path, commit or version where applicable, timestamp, component/topic tags, and freshness metadata. Verification: published metadata schema, sample exports, automated validation results, and repository-based test reports. Retrieval quality on a public Filecoin evaluation dataset Target: publish an evaluation set of Filecoin technical questions and measure the percentage of answers that retrieve relevant source-backed context. Verification: public evaluation dataset, test results, retrieval reports, and examples of source-grounded answers. Freshness tracking coverage Target: track freshness for indexed repositories and documentation sources, including last indexed date, source version, and detected updates where available. Verification: public freshness manifest, changelog, update feed, and GitHub-based indexing reports. Number of documentation gaps or outdated content issues identified Target: identify and report outdated, missing, or inconsistent Filecoin knowledge discovered during indexing and evaluation. Verification: public GitHub issues, pull requests, documentation gap reports, or issue summaries shared in the final grant report. Number of external users, teams, or AI/RAG integrations testing the knowledge base Target: at least 3 Filecoin ecosystem users, teams, builders, or AI/RAG workflows test or review the knowledge base during the grant period. Verification: public or consent-based feedback, GitHub issues, integration examples, testimonials, or final report su …[truncated] 3.3 References David Gasquez, Michael Madoff, Amin Bensadok, Marta Piekarska- Greater, Paweł Burgchardt, Duncan Ross, 4.1 Monthly Operating Burn [ "$10-$100K (small team)" ] 4.2 What % of total team monthly burn depends on this grant? Answer: 100% of the dedicated project team burn The dedicated project team for this grant would be assembled specifically to deliver the proposed scope. For that project team, 100% of the monthly burn during the grant period would depend on this grant. This does not represent 100% of the broader organization’s monthly operating burn. The broader organization will continue supporting other Filecoin-related work separately. The grant would fund the dedicated product and engineering team responsible for building the machine-readable knowledge layer, indexing pipeline, retrieval API, freshness tracking, evaluation framework, documentation, and public release. The estimated monthly cost of the dedicated project team is approximately $16,000 per month. The requested grant amount of $92,000 covers a six-month delivery period, including engineering, integrations, open-source release management, documentation, testing with Storage Providers, and community feedback cycles. 4.3 If this grant is not awarded, what happens? We will loose the opportunity to onboard for Filecoin Network very usable solution. 4.4 Core Team Mateusz Motyl – Senior AI & Full-Stack Developer A blockchain and fintech expert who already intimately understands the ecosystem's needs and challenges, having successfully built an LLM dedicated specifically to the Filecoin network. Łukasz Wal – DevOps Engineer A dedicated infrastructure specialist ensuring reliable, secure, and highly available deployments for seamless project execution. - 0.2 FTE Damian Pudło – Senior Analyst & Compliance Specialist An analytical expert bringing invaluable, hands-on ecosystem knowledge and regulatory insight from hisactive, ongoing role in the Filecoin-based FIDL project. - 0.2 FTE Marek Foss – Senior Blockchain Architect A Web3 veteran with an extensive track record of designing secure, scalable, and decentralized infrastructure across multiple prominent blockchain protocols. 0.5 FTE **Team Background** Our team has been involved in the Filecoin ecosystem for approximately three years. During this time, we have worked with Filecoin-related infrastructure, tooling, documentation, Storage Provider workflows, and AI-assisted technical knowledge systems. We have already built a proof-of-concept AI assistant that indexes Filecoin repositories, source code, and technical documentation. The current system can answer Filecoin-specific technical questions by reasoning across indexed documentation, source code, local repositories, and live project metadata. The grant will allow us to turn this proof-of-concept into a public, maintained, open-source machine-readable knowledge layer for the Filecoin ecosystem. **Why This Team?** - We are well positioned to deliver this project because: - We already have a working proof-of-concept. - We have hands-on Filecoin ecosystem experience. - We understand the fragmentation of Filecoin technical knowledge. - We have experience with AI retrieval systems, embeddings, repository indexing, and LLM interfaces. - We can combine source-code-aware indexing with Filecoin-specific domain knowledge. - We are committed to making the project open source and reusable by the ecosystem. - We are focused on practical knowledge infrastructure rather than generic documentation. 4.5 Has your team received a ProPGF grant or funding from PLFIF before? [ "No" ] 5.1 Key risks & dependencies **Risk 1: Knowledge freshness** Filecoin changes quickly. Repositories, documentation, APIs, releases, and operational practices can become outdated. Mitigation: The system will include freshness tracking, source metadata, repository monitoring, release tracking, and update workflows. Retrieval results will include source context where available, and outdated content can be flagged for review. **Risk 2: AI hallucination or incorrect answers** AI systems can produce incorrect answers if retrieval context is incomplete or stale. Mitigation: The project focuses on source-backed retrieval rather than unsupported generation. Answers and retrieved context will include source metadata. The evaluation framework will test retrieval quality and identify weak or unsupported answers. **Risk 3: Scope creep across the ecosystem** Filecoin has many repositories, products, tools, and teams. Attempting to cover everything equally could create scope risk. Mitigation: The first grant period will focus on a prioritized source list, key documentation, important repositories, and high-value technical areas. The public roadmap will clearly distinguish included sources, planned sources, and out-of-scope sources. **Risk 4: Low adoption** The knowledge base must be useful to real developers and ecosystem teams. Mitigation: The project includes community validation, example integrations, public documentation, and testing with at least 3 ecosystem users or teams. Feedback will guide prioritization and improvements. **Risk 5: Licensing and source usage** Some sources may have licensing constraints or unclear reuse permissions. Mitigation: The project will prioritize public open-source repositories and public documentation. Source metadata will be preserved, and licensing considerations will be documented. The system will avoid republishing restricted content in inappropriate ways. **Risk 6: Maintaining quality over time** A knowledge base loses value if it is not maintained after the grant period. Mitigation: The project will include a sustainability and maintenance plan, open contribution process, automated update workflows, and documentation for future maintainers. Additional support may come from community contributions, future ProPGF / RetroPGF funding, or service work for ecosystem teams that want custom integrations. Anything else you want to share that we didn't ask? **Additional Context** This project is designed as public-good knowledge infrastructure for the Filecoin ecosystem. As AI assistants become part of developer workflows, the quality of machine-readable ecosystem knowledge becomes increasingly important. Filecoin needs accurate, current, source-backed knowledge that can be consumed by both humans and machines. The project will make Filecoin technical knowledge easier to index, retrieve, validate, and reuse. It will help developers, Storage Providers, application builders, and ecosystem teams work with Filecoin more effectively. **Summary** The project will deliver an open-source, machine-readable Filecoin ecosystem knowledge layer. It will ingest and structure Filecoin documentation, repositories, source code, issues, releases, and selected operational materials. It will expose this knowledge through AI-consumable formats, retrieval APIs, metadata, evaluation tools, and public documentation. The result will be a reusable knowledge infrastructure layer that helps humans and AI systems access accurate Filecoin context. **Expected Impact** The expected impact of this project is to improve the accessibility, accuracy, and machine-readability of Filecoin technical knowledge. In the short term, the project will provide a public repository, indexed knowledge base, retrieval interface, and documentation. In the medium term, it will help developers and AI systems answer Filecoin technical questions with better source grounding and freshness context. In the long term, it can become a shared ecosystem knowledge layer used by documentation teams, pods, developers, Storage Providers, AI assistants, and operational copilots. Sustainability After the Grant After the grant, the project can continue as an open-source public-good maintained by our team and the Filecoin community. Future sustainability options may include: Community contributions. Additional ProPGF or RetroPGF funding based on demonstrated impact. Paid support or custom integrations for ecosystem teams. Integration with documentation workflows. Integration with developer portals. Integration with AI assistants or operational copilots. Public evaluation and benchmarking of Filecoin AI knowledge quality. The core project will remain open source. ****Public Application Note**** We understand that this application may be public-facing. The submitted information does not include sensitive operational details, credentials, private infrastructure data, or confidential customer information. Contributing to Core Infrastructure? Yes. The project contributes to Filecoin core infrastructure by creating a machine-readable knowledge layer for the ecosystem. Filecoin knowledge is currently distributed across many sources: Documentation websites. GitHub repositories. Source code. Specifications. Issues and pull requests. Release notes. Operational guides. Community discussions. Storage Provider workflows. Product-specific documentation. API references. Implementation details. This fragmentation makes it difficult for developers, Storage Providers, builders, and AI systems to access accurate and current information. The proposed project will create and maintain a structured knowledge layer that improves access to Filecoin technical knowledge and makes it usable by: Human developers. Storage Providers. Application builders. Ecosystem teams. Documentation maintainers. AI assistants. RAG systems. Coding agents. Operational copilots. The project supports the broader Filecoin ecosystem by improving technical discoverability, reducing repeated support questions, improving AI-assisted development, and making Filecoin knowledge easier to consume across tools and teams. Objective 1 Indirect Objective 2 Indirect Objective 3 Indirect Open Source Context N/A