{ "schemaVersion": 1, "id": "roadmap:awesome-llm-apps-roadmap", "slug": "awesome-llm-apps-roadmap", "title": "Aura Knowledge Roadmap", "summary": "A product roadmap for turning Aura Knowledge into an agent-auditable publishing system, informed by runnable examples in Shubhamsaboo/awesome-llm-apps.", "status": "published", "updatedAt": "2026-06-17", "sourceRepo": { "name": "Shubhamsaboo/awesome-llm-apps", "url": "https://github.com/Shubhamsaboo/awesome-llm-apps", "commit": "ca3a3d3", "commitDate": "2026-06-14", "reviewedAt": "2026-06-17" }, "thesis": "The knowledge garden should evolve from static article plus agent packet into a validated research object pipeline: source intake, claim verification, deterministic artifacts, reviewer gates, and optional agent-native authoring tools.", "priorityCounts": { "P0": 7, "P1": 7, "P2": 2 }, "phases": [ { "id": "phase-0", "name": "Make Evidence Packets Real", "horizon": "next foundation pass", "outcome": "Every public claim has structured support, counterevidence, source quality, and validation rules that fail weak packets before publish.", "ideaIds": [ "idea-claim-evidence-packet-v2", "idea-evidence-ci-diagnostics", "idea-claim-verification-workbench", "idea-ux-governance-checks" ] }, { "id": "phase-1", "name": "Build The Authoring Workspace", "horizon": "near-term author workflow", "outcome": "Research plans, source captures, article drafts, artifact diffs, and reviewer results become durable files rather than disappearing in chat history.", "ideaIds": [ "idea-research-workspace-builder", "idea-artifact-widget-catalog", "idea-trust-gated-publishing-pipeline" ] }, { "id": "phase-2", "name": "Expand Intake And Retrieval", "horizon": "after first public launch", "outcome": "The garden can scout new sources, import papers and URLs, suggest graph relations, and expose tool-safe retrieval paths for agents.", "ideaIds": [ "idea-always-on-source-ledger-scout", "idea-provenance-graph-v2", "idea-source-ledger-importer", "idea-corrective-research-assistant", "idea-garden-mcp-router" ] }, { "id": "phase-3", "name": "Optimize Agent Consumption", "horizon": "later workflow layer", "outcome": "Agents can consume compact packets, inspect evidence interactively, and regression-test briefs without turning the public site into a runtime app.", "ideaIds": [ "idea-compact-agent-feeds", "idea-claim-evidence-canvas", "idea-agent-brief-evaluation-harness", "idea-agent-native-packet-inspector" ] } ], "ideas": [ { "id": "idea-claim-evidence-packet-v2", "title": "Claim Evidence Packet v2", "priority": "P0", "category": "evidence", "sourcePaths": [ "rag_tutorials/knowledge_graph_rag_citations/README.md", "rag_tutorials/multimodal_agentic_rag/README.md", "advanced_ai_agents/single_agent_apps/earnings_call_analyst_agent/schemas.py" ], "pattern": "Retrieval results and citation UI are backed by typed objects that carry source, evidence, confidence, and provenance.", "gardenMapping": "Extend claim records beyond source IDs into evidence snippets, source locations, support type, counterevidence, extraction metadata, and reviewer status.", "firstImplementation": "Add an evidencePackets array to artifact.json and render one compact evidence card per claim in the article audit section.", "risks": [ "Citation IDs can become theater if snippets do not actually support the claim.", "Evidence snippets must be short and attributable without copying long copyrighted passages." ] }, { "id": "idea-evidence-ci-diagnostics", "title": "Evidence CI And RAG Diagnostics", "priority": "P0", "category": "governance", "sourcePaths": [ "rag_tutorials/rag_failure_diagnostics_clinic/README.md", "rag_tutorials/corrective_rag/README.md", "rag_tutorials/contextualai_rag_agent/README.md" ], "pattern": "A small failure taxonomy, relevance grading, and answer evaluation can catch structural problems earlier than prompt tweaks.", "gardenMapping": "Fail CI for orphan claims, empty evidence, missing counterevidence, stale source dates, low source diversity, and dangling graph edges.", "firstImplementation": "Add deterministic validators first, then allow optional LLM audit reports as non-blocking artifacts until the taxonomy is stable.", "risks": [ "LLM audits can be noisy and should not replace deterministic schema checks.", "External link checks can be flaky and should use retries or scheduled mode." ] }, { "id": "idea-claim-verification-workbench", "title": "Claim Verification Workbench", "priority": "P0", "category": "evidence", "sourcePaths": [ "awesome_agent_skills/fact-checker/SKILL.md", "awesome_agent_skills/deep-research/SKILL.md", "awesome_agent_skills/academic-researcher/SKILL.md", "advanced_ai_agents/single_agent_apps/ai_deep_research_agent/README.md" ], "pattern": "Research agents extract claims, define required evidence, rate source quality, and synthesize findings with citations.", "gardenMapping": "Create an author-side workflow that walks each claim through required evidence, source quality, gap notes, counterclaims, and last-checked date.", "firstImplementation": "Add verification fields to claim objects and a report command that lists claims by verified, needs evidence, contested, or stale.", "risks": [ "Source quality ratings need human review for contested or interpretive claims.", "The workbench should help authors decide; it should not silently approve claims." ] }, { "id": "idea-ux-governance-checks", "title": "UX Governance Checks", "priority": "P0", "category": "interface", "sourcePaths": [ "awesome_agent_skills/ux-designer/SKILL.md" ], "pattern": "UX review can be encoded as reusable criteria for accessibility, hierarchy, interaction, and information architecture.", "gardenMapping": "Make reading quality part of the publishing gate: mobile flow, keyboard access, source ledger scanability, claim marker clarity, and dark mode.", "firstImplementation": "Add a docs checklist and optional Playwright assertions for no horizontal overflow, collapsed mobile audit rail, visible focus, and route availability.", "risks": [ "Visual polish can drift into dashboard creep if every data point becomes a widget.", "Automated checks cannot judge whether the article actually reads well." ] }, { "id": "idea-research-workspace-builder", "title": "Research Workspace Builder", "priority": "P0", "category": "authoring", "sourcePaths": [ "generative_ui_agents/ai-deep-research-agent/README.md", "advanced_ai_agents/single_agent_apps/research_agent_gemini_interaction_api/README.md" ], "pattern": "A sidecar workspace captures plans, todos, files, research runs, and synthesis outputs as the agent works.", "gardenMapping": "Create a local authoring mode where research plans, source captures, claim packets, and final article artifacts are durable files in the repo.", "firstImplementation": "Start with CLI scaffolding for an article workspace before adding any live UI.", "risks": [ "A runtime authoring app is larger than the static site and should remain local until the workflow proves itself.", "Chat history should not become the source of truth." ] }, { "id": "idea-artifact-widget-catalog", "title": "Artifact Widget Catalog", "priority": "P0", "category": "interface", "sourcePaths": [ "generative_ui_agents/generative-ui-starter-project/README.md", "generative_ui_agents/ai-shadcn-component-generator/README.md" ], "pattern": "Generative UI works best when agents choose from bounded, typed components rather than arbitrary executable UI.", "gardenMapping": "Define approved static components for ClaimCard, SourceLedger, MaturityBadge, StemGraph, AgentPacketPreview, and RoadmapItem.", "firstImplementation": "Document component contracts and render the roadmap page from a schema-validated JSON artifact.", "risks": [ "Unbounded generated UI is a security and maintenance risk for a public static site.", "Widget schemas must stay aligned with article and roadmap schemas." ] }, { "id": "idea-trust-gated-publishing-pipeline", "title": "Trust-Gated Publishing Pipeline", "priority": "P0", "category": "governance", "sourcePaths": [ "advanced_ai_agents/multi_agent_apps/trust_gated_agent_team/README.md", "advanced_ai_agents/multi_agent_apps/multi_agent_trust_layer/README.md" ], "pattern": "Agent participation is gated by trust checks, scoped roles, and hash-chained audit records.", "gardenMapping": "Record agent roles, inputs, output hashes, review status, policy scope, and human approval before an article can move to published.", "firstImplementation": "Add optional provenance.json per article and make published status require a humanReview block plus generated artifact diff.", "risks": [ "Hash chains alone are not a complete security model without signing or trusted timestamps.", "Sensitive prompts and private notes should be hashed or summarized, not published." ] }, { "id": "idea-always-on-source-ledger-scout", "title": "Always-On Source Ledger Scout", "priority": "P1", "category": "authoring", "sourcePaths": [ "always_on_agents/always_on_hn_briefing_agent/README.md", "advanced_ai_agents/multi_agent_apps/devpulse_ai/README.md" ], "pattern": "Scheduled agents collect, rank, normalize, and deliver high-signal briefs with dry-run delivery by default.", "gardenMapping": "Monitor papers, product docs, protocols, GitHub repos, and newsletters for new evidence, counterevidence, and article seeds.", "firstImplementation": "Create a dry-run scout that writes candidate source-ledger entries and article issue drafts without sending or publishing anything.", "risks": [ "Always-on collection can create noise without clear scoring and deduplication.", "Live browsing sources must be normalized before they support a public claim." ] }, { "id": "idea-provenance-graph-v2", "title": "Provenance Graph v2", "priority": "P1", "category": "retrieval", "sourcePaths": [ "rag_tutorials/knowledge_graph_rag_citations/README.md" ], "pattern": "Knowledge graph RAG uses entity and relation extraction with multi-hop reasoning traces and source attribution.", "gardenMapping": "Expand the graph from topic/article/claim/source into concept, entity, method, supports, contests, depends-on, mentions, and derived-from edges.", "firstImplementation": "Add reviewed edge metadata and show why each public graph edge exists.", "risks": [ "LLM-extracted edges can invent relation labels or merge distinct concepts.", "Graph visuals can create false confidence if provenance is hidden." ] }, { "id": "idea-source-ledger-importer", "title": "Source Ledger Importer", "priority": "P1", "category": "authoring", "sourcePaths": [ "advanced_llm_apps/chat_with_X_tutorials/chat_with_research_papers/README.md", "advanced_llm_apps/llm_apps_with_memory_tutorials/ai_arxiv_agent_memory/README.md", "advanced_llm_apps/chat_with_X_tutorials/chat_with_substack/README.md" ], "pattern": "Paper, newsletter, and URL chat examples show how to ingest external corpora for retrieval and summarization.", "gardenMapping": "Add a source CLI that imports arXiv, DOI, URL, GitHub, and newsletter items into the ledger with normalized metadata and candidate claim links.", "firstImplementation": "Start with URL and arXiv metadata import, then require manual approval before sources can support claims.", "risks": [ "Personal memory should stay private and not leak into public artifacts.", "Imported summaries must be treated as draft notes, not evidence by themselves." ] }, { "id": "idea-corrective-research-assistant", "title": "Corrective Research Assistant", "priority": "P1", "category": "evidence", "sourcePaths": [ "rag_tutorials/corrective_rag/README.md", "advanced_ai_agents/multi_agent_apps/ai_domain_deep_research_agent/README.md", "rag_tutorials/ai_blog_search/README.md" ], "pattern": "Agentic RAG can grade relevance, rewrite weak queries, retrieve more context, and identify missing evidence.", "gardenMapping": "Audit drafts for missing sources, weak counterarguments, ungrounded claims, and unclear research questions before publication.", "firstImplementation": "Add a non-blocking draft audit report that groups suggestions by claim ID and severity.", "risks": [ "Query rewrite decisions need observability so agents do not silently chase a different question.", "The assistant should produce suggestions, not rewrite the article automatically." ] }, { "id": "idea-garden-mcp-router", "title": "Garden MCP Router", "priority": "P1", "category": "retrieval", "sourcePaths": [ "mcp_ai_agents/multi_mcp_agent_router/README.md", "mcp_ai_agents/github_mcp_agent/README.md", "mcp_ai_agents/browser_mcp_agent/README.md", "mcp_ai_agents/notion_mcp_agent/README.md" ], "pattern": "A router sends requests to specialist agents with only the MCP tools needed for that task.", "gardenMapping": "Expose safe garden operations as tools: query artifacts, inspect graph slices, validate packets, create scaffolds, update source ledgers, and open PRs.", "firstImplementation": "Define tool contracts and dry-run semantics before adding any write-capable MCP server.", "risks": [ "Over-permissioned MCP tools can mutate repos or browse accounts unexpectedly.", "Write operations need explicit dry-run previews and human approval." ] }, { "id": "idea-compact-agent-feeds", "title": "Compact Agent Feeds", "priority": "P1", "category": "optimization", "sourcePaths": [ "advanced_llm_apps/llm_optimization_tools/toonify_token_optimization/README.md", "advanced_llm_apps/llm_optimization_tools/headroom_context_optimization/README.md" ], "pattern": "Compact serializations and context compression reduce redundant structured data while preserving retrieval hooks.", "gardenMapping": "Generate derived compact feeds for claims, sources, roadmap items, and graph edges beside canonical JSON and JSONL.", "firstImplementation": "Add token estimates to agent feeds and experiment with a compact TSV or TOON-style claims table as a derived artifact.", "risks": [ "Canonical artifacts should remain JSON; compact packets must be derived and hash-linked.", "Compression must not remove counterevidence or maturity markers." ] }, { "id": "idea-claim-evidence-canvas", "title": "Claim Evidence Canvas", "priority": "P1", "category": "interface", "sourcePaths": [ "generative_ui_agents/ai-dashboard-canvas-agent/README.md" ], "pattern": "Agents populate addressable dashboard state, charts, and panels instead of returning only text.", "gardenMapping": "Add an optional evidence view showing claim coverage, source diversity, maturity, counterevidence, and graph health from artifacts.", "firstImplementation": "Start as a static summary on the roadmap or graph page, not a runtime dashboard.", "risks": [ "Dashboard creep can distract from focused reading.", "Every metric must trace back to artifact.json or a source-ledger entry." ] }, { "id": "idea-agent-brief-evaluation-harness", "title": "Agent Brief Evaluation Harness", "priority": "P2", "category": "optimization", "sourcePaths": [ "awesome_agent_skills/self-improving-agent-skills/README.md" ], "pattern": "A multi-agent loop generates scenarios, scores outputs against criteria, diagnoses failures, and keeps only targeted improvements.", "gardenMapping": "Regression-test agent.md and generated packets against expected retrieval queries and citation behavior.", "firstImplementation": "Create a small static eval set before attempting self-improvement.", "risks": [ "Optimization loops can overfit narrow evals.", "Human prose quality should not be optimized by an automated loop." ] }, { "id": "idea-agent-native-packet-inspector", "title": "Agent-Native Packet Inspector", "priority": "P2", "category": "interface", "sourcePaths": [ "generative_ui_agents/mcp-apps-generative-ui-showcase/README.md", "generative_ui_agents/ai-mcp-app-builder/README.md" ], "pattern": "MCP app examples link tool calls to sandboxed interactive resources and generated apps.", "gardenMapping": "Expose a local or server-side inspector where agents request graph slices, packet summaries, or review reports from garden artifacts.", "firstImplementation": "Keep this out of the public GitHub Pages site until the static artifact contract stabilizes.", "risks": [ "Runtime agent apps add hosting, security, and authentication concerns.", "Sandboxed tools should never write to main without PR review." ] } ], "principles": [ "Keep the public site static, readable, and inspectable; move runtime agent features into local authoring or separate services.", "Use deterministic validation for schemas, paths, hashes, generated files, and graph integrity before adding LLM judgment.", "Treat source-ledger entries as the only path from external material to public claims.", "Make every agent-facing compression or UI surface a derived artifact with a canonical JSON source.", "Prefer dry-run previews and PR-based workflows for any write-capable automation.", "Keep personal research memory private unless a specific note is intentionally promoted into a public source or article artifact." ], "crossAgentReview": { "status": "reviewed-with-agent-input", "reviewedAt": "2026-06-17", "reviewers": [ "Codex main agent", "Ohm RAG and memory reviewer", "Bacon agent workflow reviewer", "Gauss generative UI reviewer" ], "notes": "Three independent research passes reviewed RAG/research memory examples, agent/MCP workflows, and generative UI/token-efficiency examples. The roadmap prioritizes durable evidence and governance before runtime agent interfaces." }, "agentInstructions": [ "Use idea IDs as retrieval units.", "Treat P0 as next foundation work, P1 as near-term roadmap, and P2 as later experiment.", "Prefer deterministic validators over agent judgment when a check can be expressed structurally.", "Do not infer that every source example should be adopted as a dependency; most are pattern references." ], "generatedAt": "2026-06-29T00:00:00.000Z", "pageUrl": "https://aura-knowledge.github.io/roadmap/", "agentJsonPath": "/agents/roadmap/awesome-llm-apps-roadmap.json", "sourceRepoPath": "content/roadmap/awesome-llm-apps-roadmap.json", "sourceGitHubUrl": "https://github.com/aura-knowledge/aura-knowledge.github.io/blob/main/content/roadmap/awesome-llm-apps-roadmap.json" }