Name: GD Chatbot v2 Tagline: AI-powered Grateful Dead historian and music discovery assistant with RAG, web search, streaming responses, and multi-platform music integration. Current status: Live First commit / project start: V1 (gd-claude-chatbot) archived February 2026 per archive/README.md (reached v1.7.1–v1.9.5); v2 rebuild activity runs ~January 2026 per CHANGELOG (earliest entry 2.0.4 dated 2026-01-12); v2.2.0 implementation completed February 12, 2026 per IMPLEMENTATION-SUMMARY.md. No per-file git history available — repository first committed to ITI monorepo 2026-03-27 as a workspace-wide pre-migration snapshot.
GD Chatbot v2 is a WordPress plugin that creates a domain-specific AI assistant for the Grateful Dead community. Powered by Claude, it combines a comprehensive bundled knowledge base (band history, music catalog, equipment, culture, art galleries, literature, setlists for 2,340+ shows from 1965–1995), optional Pinecone vector RAG, optional Tavily web search, and multi-platform music discovery (Archive.org, Spotify, Apple Music, YouTube Music, Amazon Music, Tidal). The chatbot features intent-based context assembly with token budget management, streaming responses via SSE, song detection in assistant text with clickable playback links, and themed UI options (psychedelic GD theme or professional theme). It serves as both a knowledge resource for Deadheads and a pattern for domain-specific RAG chatbot development.
Three capabilities distinguish GD Chatbot v2. First, its deep domain knowledge base — 8 core knowledge documents, 21 supplementary deep-dive files, 3 disambiguation documents, and 31 years of setlist CSVs (1965–1995) — creates an information density that generic AI chatbots cannot match for Grateful Dead queries. Second, the hybrid music layer detects song mentions in assistant responses and surfaces clickable links to Archive.org recordings plus 5 commercial streaming services, turning knowledge into immediate listening experiences. Third, the RAG hygiene discipline — strict guardrails prevent the AI from revealing internal retrieval mechanics, inventing venue locations, or contradicting the authoritative knowledge base (e.g., The Bahr Gallery must always be “Oyster Bay, Long Island, NY”).
Platform: WordPress plugin (PHP 7.4+, WordPress 5.0+) Deployment: Self-hosted WordPress, single-site Primary interface: Shortcode Your guide to all things Grateful Dead as inline embed or floating widget; themed UI (psychedelic GD default, professional optional)
GD Chatbot
Primary user: Deadheads and Grateful Dead enthusiasts seeking accurate, deep knowledge about the band’s history, music, recordings, equipment, and culture Secondary users: WordPress site owners deploying a Grateful Dead knowledge resource; developers adapting the domain-specific RAG pattern for other communities; music archive explorers User environment: Grateful Dead community websites running WordPress; inline chatbot or floating widget on any page
The Grateful Dead’s 30+ year history encompasses thousands of live shows, hundreds of songs with complex performance histories, evolving equipment rigs, a rich cultural ecosystem (art, literature, venues, community), and a massive archive of live recordings. This information is scattered across dozens of websites, books, databases, and personal knowledge. Generic AI chatbots can answer basic questions but lack the depth, accuracy, and citation specificity that knowledgeable Deadheads expect — and they certainly cannot connect knowledge to immediate listening experiences across Archive.org and streaming platforms.
| Need | How the product addresses it | Source of evidence |
|---|---|---|
| Deep, accurate Grateful Dead knowledge | Bundled KB: 8 core docs, 21 supplementary files, 31 years of setlist CSVs (2,340+ shows) | plugin/context/ directory structure |
| Setlist and show intelligence | CSV-based setlist search triggered by show/setlist query intent detection | class-query-optimizer.php, setlists/*.csv |
| Knowledge-to-listening bridge | Song detection in responses + modal with Archive.org playback + 5 streaming service links | class-response-enricher.php, class-streaming-service-manager.php |
| Multi-source retrieval | Bundled KB + optional Pinecone + optional Tavily + Knowledgebase Loader + AI Power plugin integration | class-chat-handler.php multi-source assembly |
| Token-efficient context | Intent-based context selection with token budget management and context caching | class-query-optimizer.php, class-context-builder.php, token budget manager |
Deadheads consulted Deadbase, setlist databases (dead.net, setlists.net), Archive.org directly, forum posts, books, and personal memory. Finding authoritative information required cross-referencing multiple sources. Connecting knowledge to actual recordings meant separately searching Archive.org or streaming platforms. Generic AI chatbots could provide surface-level answers but lacked the depth and accuracy that the community demands and could not distinguish between authoritative and uncertain information.
Primary category: Domain-specific AI knowledge assistant / music community chatbot Market maturity: Niche — AI chatbots for specific music communities are rare; the Grateful Dead community is unusually well-documented but underserved by AI tools Key dynamics: The Grateful Dead community is uniquely suited to AI-assisted knowledge: massive documentation, passionate users who value accuracy, and a culture of sharing (tape trading, Archive.org). The live recording archive creates a direct knowledge-to-experience bridge that most AI chatbots cannot offer. [CLAUDE NOTE: inferred from community characteristics]
| Product / Company | Approach | Strengths | Key gap this project addresses | Source |
|---|---|---|---|---|
| dead.net / Grateful Dead official | Official band site | Authoritative source, show archives | No AI chatbot, limited interactive search | ⚡ General market knowledge |
| Archive.org Grateful Dead collection | Recording archive | 15,000+ live recordings | Browse/search interface, no conversational knowledge | ⚡ General market knowledge |
| Generic AI chatbots (ChatGPT, Claude.ai) | General knowledge | Broad capabilities | Shallow GD knowledge, no setlist intelligence, no recording links, no accuracy guardrails | ⚡ General market knowledge |
| Deadbase / setlist databases | Structured data | Comprehensive setlist data | Data-only, no conversational interface, no contextual knowledge | ⚡ General market knowledge |
GD Chatbot v2 occupies a unique position as the only AI-powered conversational interface purpose-built for the Grateful Dead community. It combines the depth of specialized databases (setlists, equipment, venues) with the conversational fluency of modern LLMs and the music discovery capability of streaming platform integration — creating a “know it, hear it” experience that no existing tool provides. [CLAUDE NOTE: inferred from product capabilities]
The product’s moat has multiple layers: (1) curated knowledge base with deep supplementary content (interviews, academic papers, dissertations, gallery guides, equipment specifications) that would take significant effort to replicate; (2) 31 years of structured setlist data; (3) multi-platform music integration (Archive.org + 5 streaming services) with song detection; (4) accuracy guardrails developed from domain-specific knowledge that generic AI lacks; (5) the plugin serves as a reusable pattern for domain-specific RAG chatbots, creating ecosystem value. The CLAUDE.md notes derivatives including ITI Chatbot, AI News Cafe, and Scuba GPT.
The product evolved from a v1 chatbot (gd-claude-chatbot, now archived) to a full v2 rebuild with improved architecture, streaming, token optimization, and music integration. Requirements were driven by community knowledge depth, accuracy expectations, and the insight that connecting knowledge to listening experiences creates unique value. [CLAUDE NOTE: inferred from archive/ directory and version history]
Hard constraints:
Explicit non-goals:
| Decision | Alternatives considered | Rationale | Evidence source |
|---|---|---|---|
| v2 as separate plugin (gd_chatbot_v2_ prefix) | In-place upgrade of v1 | Side-by-side installation enables migration; separate options namespace prevents conflicts | plugin/CHANGELOG.md, gd_chatbot_v2_ option prefix |
| Token optimization mode | Full KB injection every request | Manages costs by loading only intent-relevant context fragments | class-query-optimizer.php, class-context-builder.php |
| Hardcoded accuracy guardrails in PHP | Prompt-only guardrails | Critical accuracy rules (Bahr Gallery location, no RAG disclosure) cannot be accidentally modified via admin prompt editing | class-claude-api.php guardrail strings |
| Song detection in responses | Manual song linking | Seamless knowledge-to-listening experience; songs mentioned in AI responses become immediately listenable | class-response-enricher.php |
| Multi-platform streaming via OAuth | Single platform (e.g., Spotify only) | Maximizes audience reach; users have different platform preferences | class-streaming-service-manager.php, docs/ADMIN-GUIDE-STREAMING-SERVICES.md |
| CSV-based setlist storage | Database-only | Portable, version-controllable, human-readable; loaded on-demand by query optimizer | plugin/context/setlists/*.csv |
KB type: Bundled markdown + CSV with optional Pinecone vector augmentation and optional Tavily web search Location in repo: plugin/context/ (core, supplementary, disambiguation, reference, setlists) Estimated size: ~60+ files including 31 setlist CSVs, 8 core docs, 21 supplementary files, 3 disambiguation docs, 3 reference CSVs
plugin/context/
├── core/ # 8 primary knowledge documents
│ ├── band-and-history.md
│ ├── music-and-recordings.md
│ ├── equipment.md
│ ├── culture-and-community.md
│ ├── galleries-and-art.md
│ ├── books-and-literature.md
│ ├── resources-and-media.md
│ └── terminology.md
├── supplementary/ # 21 deep-dive documents
│ ├── interviews, UCSC archive, academic papers
│ ├── dissertations, gear guides, statistics
│ └── ... (more specialized topics)
├── disambiguation/ # 3 disambiguation guides
│ ├── song-titles.md
│ ├── duplicate-titles.md
│ └── equipment-names.md
├── reference/ # 3 structured data CSVs
│ ├── songs.csv
│ ├── equipment.csv
│ └── tavily-domains.csv # Domain allowlist for web search
├── setlists/ # 31 yearly CSVs (1965–1995)
│ ├── 1965.csv
│ ├── ...
│ └── 1995.csv
└── _archive/ # Dev/planning — NOT loaded at runtime
| Category | Files / format | Purpose | Update frequency |
|---|---|---|---|
| Core knowledge | 8 × .md in core/ | Band history, music catalog, equipment, culture, galleries, books, resources, terminology | Per release |
| Deep dives | 21 × .md in supplementary/ | Interviews, academic papers, dissertations, gear specifications, statistics, gallery guides | Per release |
| Disambiguation | 3 × .md in disambiguation/ | Song title disambiguation, duplicate title resolution, equipment name clarification | Per release |
| Structured reference | 3 × .csv in reference/ | Songs catalog, equipment catalog, Tavily domain allowlist | Per release |
| Setlist data | 31 × .csv in setlists/ (1965–1995) | Show-by-show setlist data, 2,340+ shows | Historical (fixed) |
| Bahr Gallery (authoritative) | bahr-gallery.md in supplementary/ | Single source of truth for Bahr Gallery (Oyster Bay, Long Island, NY); overrides other KB mentions | Per release |
The knowledge base was built through systematic documentation of Grateful Dead domains: band history and timeline, complete discography and recording catalog, equipment evolution, cultural ecosystem, galleries and art, and literature. Setlist CSVs covering 31 years (1965–1995) were compiled from historical show records. Supplementary documents provide academic and interview depth. The disambiguation guides address known confusion points (songs with similar titles, equipment name variations). Reference CSVs provide structured lookup data for songs and equipment. The Tavily domain allowlist curates trusted web sources for live search augmentation. [CLAUDE NOTE: inferred from KB structure and guardrail logic]
System prompt approach: Admin-editable WordPress option gd_chatbot_v2_claude_system_prompt with default set on activation via get_default_system_prompt() (Grateful Dead historian role, tone, accuracy/formatting rules). Runtime appends vary by mode: Full KB mode loads all core + supplementary + disambiguation + long guardrails; Optimized mode uses short guardrails + intent-driven fragments via GD_Context_Builder. Key behavioural guardrails: Do not disclose internal RAG mechanics (Pinecone, Tavily, KB source); strict venue/business location accuracy; no fabricated show dates or setlists. Persona / tone configuration: Grateful Dead historian — knowledgeable, enthusiastic, accurate; Deadhead community voice. Tool use / function calling: No LLM tool use — context assembly is server-side. Song detection and enrichment happens post-response in class-response-enricher.php.
GD Chatbot v2 is a ground-up rebuild of the original gd-claude-chatbot (v1.7.1, now archived). The v2 architecture uses separate namespacing ( Your guide to all things Grateful Deadgd_chatbot_v2_ options, shortcode) enabling side-by-side installation during migration. Development progressed from core chatbot functionality through token optimization, streaming, and culminated in the hybrid music integration layer. The IMPLEMENTATION-SUMMARY.md documents approximately 30 hours of implementation work completed by February 12, 2026.
GD Chatbot
| Phase | Approximate timeframe | What was built | Key commits or milestones |
|---|---|---|---|
| v1.x (archived) | Pre-2026 [CLAUDE NOTE: inferred] | Original gd-claude-chatbot through v1.7.1 | Archived in archive/ directory |
| v2.0.0–2.0.6 | ~January 2026 | Core v2 rebuild: new architecture, bundled KB, token optimization, streaming, Claude API integration, themed UI | CHANGELOG entries 2.0.0–2.0.6 |
| v2.1.0 | ~February 2026 [CLAUDE NOTE: inferred] | Hybrid music layer: Archive.org integration, DB sync, song detection, clickable song links, modal UI | IMPLEMENTATION-SUMMARY.md |
| v2.2.0 | February 12, 2026 | Multi-platform streaming: OAuth for Spotify, Apple Music, YouTube Music, Amazon Music, Tidal; unified search; encrypted credentials; admin streaming dashboard | releases/README.md, docs/ADMIN-GUIDE-STREAMING-SERVICES.md |
Development uses Cursor IDE with CLAUDE.md providing comprehensive product context including knowledge base paths, architectural decisions, and ITI shared library references. The CLAUDE.md notes that the GD Chatbot pattern has been replicated in derivative products (ITI Chatbot, AI News Cafe, Scuba GPT), demonstrating the plugin’s role as an architectural reference implementation for domain-specific RAG chatbots.
| Challenge | How resolved | Evidence |
|---|---|---|
| Token cost management with large KB | Token optimization mode: GD_Query_Optimizer detects intent, GD_Context_Builder loads only relevant fragments, GD_Token_Budget_Manager caps context size | class-query-optimizer.php, class-context-builder.php |
| Accuracy for venue/business locations | Hardcoded PHP guardrails that strip conflicting KB mentions and inject authoritative source (bahr-gallery.md); cannot be bypassed via admin prompt editing | class-claude-api.php guardrail implementation |
| Knowledge-to-listening bridge | class-response-enricher.php detects song mentions in assistant text; generates clickable links to Archive.org + streaming platforms; modal UI for playback | class-response-enricher.php, song-modal.js |
| Multi-platform OAuth complexity | Encrypted credential storage per platform; streaming service manager abstracts platform differences; admin dashboard for per-service configuration | class-streaming-service-manager.php, docs/ADMIN-GUIDE-STREAMING-SERVICES.md |
| V1-to-v2 migration | Separate namespace (gd_chatbot_v2_*) enables side-by-side installation; users can run both versions during transition | CHANGELOG.md, plugin/README.md v2 migration notes |
| Context caching | GD_Context_Cache stores assembled context fragments to avoid repeated KB loading across similar queries | Context cache implementation |
| Model / API | Provider | Role in product | Integration method |
|---|---|---|---|
| Claude (default: claude-sonnet-4-20250514) | Anthropic | Conversational AI, knowledge synthesis, Grateful Dead expertise | Messages API via wp_remote_post; streaming via SSE |
| Tavily Search | Tavily | Optional real-time web search with domain allowlist | Search API with curated tavily-domains.csv |
| Pinecone | Pinecone | Optional vector semantic search across knowledge base | Vector similarity search API |
| Archive.org | Internet Archive | Live recording search and playback links (15,000+ GD recordings) | Archive.org API |
| Spotify | Spotify | Commercial streaming song search and links | OAuth + Web API |
| Apple Music | Apple | Commercial streaming song search and links | MusicKit / API |
| YouTube Music | Commercial streaming song search and links | YouTube Data API | |
| Amazon Music | Amazon | Commercial streaming song search and links | OAuth + API |
| Tidal | Tidal | Commercial streaming song search and links | OAuth + API |
| Tool | Category | Purpose |
|---|---|---|
| Chat Handler | Orchestration | Central coordinator: receives message, assembles context from multiple sources, calls Claude, enriches response |
| Query Optimizer | Intent detection | Classifies queries by topic (setlist, song, venue, equipment, history, etc.) for targeted context loading |
| Context Builder | RAG assembly | Loads intent-relevant KB fragments within token budget |
| Token Budget Manager | Cost control | Caps total context tokens to control API costs |
| Context Cache | Performance | Caches assembled context for similar queries |
| Response Enricher | Post-processing | Detects song mentions in Claude responses; generates streaming platform links |
| Streaming Service Manager | Music integration | Manages OAuth credentials and unified search across 6 platforms |
| Tool | How used in build |
|---|---|
| Cursor IDE | Primary development environment with CLAUDE.md context |
| Claude AI | Knowledge base content development, prompt engineering [CLAUDE NOTE: inferred] |
| Antigravity | Autonomous test execution, browser QA, visual regression testing — used per global CLAUDE.md tool lane |
| Version / Phase | Date | Summary of changes | Significance |
|---|---|---|---|
| v1.0–1.7.1 (archived) | Pre-2026 | Original gd-claude-chatbot through final v1 release | Foundation / now archived |
| 2.0.0–2.0.3 | ~January 2026 | Core v2 rebuild: new architecture, bundled KB, token optimization, themed UI | Architecture overhaul |
| 2.0.4 | ~January 2026 | Stability improvements | Incremental |
| 2.0.5 | ~January 2026 | Token optimization refinements | Performance |
| 2.0.6 | ~January 2026 | Bug fixes and polish | Last version documented in plugin CHANGELOG |
| 2.1.0 | ~February 2026 [CLAUDE NOTE: inferred] | Archive.org integration, song detection, modal playback UI | Music discovery launch |
| 2.2.0 | February 12, 2026 | Multi-platform streaming: Spotify, Apple Music, YouTube Music, Amazon Music, Tidal; OAuth; encrypted credentials; admin streaming dashboard | Full music integration |
The most significant pivot was from a pure knowledge chatbot (v2.0.x) to a hybrid knowledge + music discovery platform (v2.1–2.2). The addition of song detection in AI responses and multi-platform streaming integration transformed the product from “ask about the Dead” to “ask, learn, and listen” — a fundamentally different user experience. The v1-to-v2 rebuild itself was a major architectural reset, moving from the original gd-claude-chatbot to a cleanly namespaced, token-optimized, streaming-capable architecture.
gd_chatbot_streaming_enabled) uses old prefix without v2, inconsistent with namespacing convention| Artifact | Path | Type | What it shows |
|---|---|---|---|
| GD psychedelic theme | plugin/assets/css/gd-theme.css | Stylesheet | Default Grateful Dead-themed visual design |
| Professional theme | plugin/assets/css/professional-theme.css | Stylesheet | Alternative clean/corporate visual design |
| Chatbot JS | plugin/assets/js/chatbot.js | JavaScript | Chat interaction, streaming display, source attribution |
| Song modal JS | plugin/assets/js/song-modal.js | JavaScript | Song detection modal, Archive.org + streaming links |
| Document | Path | Type | Status |
|---|---|---|---|
| Plugin README | plugin/README.md | Markdown | Covers through v2.0.3 |
| WordPress readme | plugin/readme.txt | Text | Stable tag 2.0.3 |
| CHANGELOG | plugin/CHANGELOG.md | Markdown | Entries through v2.0.6 |
| Implementation Summary | docs/IMPLEMENTATION-SUMMARY.md | Markdown | Complete v2.2.0 implementation narrative (~30 hours) |
| Streaming Services Admin Guide | docs/ADMIN-GUIDE-STREAMING-SERVICES.md | Markdown | OAuth setup for 5 streaming platforms |
| Music Streaming Requirements | docs/music-streaming-requirements.md | Markdown | Requirements specification |
| Token Management Requirements | docs/token-management-requirements.md | Markdown | Token optimization requirements |
| Release notes (multiple) | docs/release-notes/ | Markdown | Per-version release narratives |
| GitHub CONTRIBUTING | .github/CONTRIBUTING.md | Markdown | Contribution guidelines |
| GitHub SECURITY | .github/SECURITY.md | Markdown | Security policy |
| Artifact | Path | Description |
|---|---|---|
| Core knowledge (8 docs) | plugin/context/core/ | Band history, music, equipment, culture, galleries, books, resources, terminology |
| Supplementary knowledge (21 docs) | plugin/context/supplementary/ | Deep dives: interviews, academic papers, gear guides, gallery guides, statistics |
| Disambiguation (3 docs) | plugin/context/disambiguation/ | Song titles, duplicate titles, equipment names |
| Setlist CSVs (31 years) | plugin/context/setlists/ | 1965–1995 show-by-show setlists, 2,340+ shows |
| Reference CSVs | plugin/context/reference/ | Songs catalog, equipment catalog, Tavily domain allowlist |
| Release ZIP (v2.2.0) | releases/ | 398KB, 125 files per releases/README.md |
| Archived v1 | archive/ | Original gd-claude-chatbot for reference |
| Build scripts | scripts/ | build-release.sh, cleanup-for-release.sh |
The product originated from the intersection of two observations: the Grateful Dead community has an unusually rich and well-documented history (thousands of shows, extensive recordings, passionate knowledge culture), and modern AI RAG techniques make it possible to build domain-specific assistants that match community knowledge expectations. The v1 chatbot (gd-claude-chatbot) proved the concept; v2 rebuilt the architecture for token efficiency, streaming, and — critically — the insight that connecting knowledge to immediate listening experiences via Archive.org and streaming platforms creates a qualitatively different product. [CLAUDE NOTE: inferred from archive/ directory and v2 feature evolution]
Research approach used: NOT FOUND — add manually Key market observations:
What existing products got wrong: Generic AI chatbots lack domain depth. Databases (setlist sites, Archive.org) provide data without conversational synthesis. No tool bridges the gap between “learn about it” and “listen to it.” [CLAUDE NOTE: inferred]
If we build a domain-specific AI assistant with deep Grateful Dead knowledge, strict accuracy guardrails, and the ability to connect any knowledge response to immediate listening via Archive.org and streaming platforms, the Deadhead community will adopt it as a primary knowledge and discovery tool — and the architecture will prove replicable for other domain-specific communities. [CLAUDE NOTE: inferred from product design and CLAUDE.md derivative product references]
The product evolved through three distinct phases: (1) v1 proof-of-concept (gd-claude-chatbot) established that domain-specific RAG chatbots work for music communities; (2) v2.0.x rebuilt the architecture with token optimization, streaming, and themed UI for production quality; (3) v2.1–2.2 added the transformative music integration layer (Archive.org + 5 streaming services with song detection). The pattern proved replicable: CLAUDE.md documents derivatives including ITI Chatbot, AI News Cafe, and Scuba GPT. The project also generated reusable documentation and requirements artifacts (music streaming requirements, token management requirements) that benefit the broader ITI product portfolio.
What works well: Deep domain knowledge base with comprehensive coverage of Grateful Dead history, music, equipment, and culture. Intent-based token optimization keeps API costs manageable. Streaming responses provide responsive UX. Song detection and multi-platform music links create a unique “learn and listen” experience. Architecture has proven replicable for derivative products. Accuracy guardrails (especially Bahr Gallery enforcement) demonstrate RAG hygiene best practices. Current limitations: Plugin CHANGELOG only covers through v2.0.6 — v2.1–2.2 features are documented separately but not in the formal changelog. Version numbers are inconsistent (plugin header 2.2.0 vs. readme.txt stable tag 2.0.3). Setlist data ends at 1995. One option key (gd_chatbot_streaming_enabled) doesn’t follow the v2 naming convention. The _archive/ directory in context/ contains dev/planning files that should not be present in distribution packages. Estimated completeness: Live (v2.2.0) — production-ready with comprehensive knowledge and music integration.
gd_chatbot_streaming_enabled option to use v2 prefix_archive/ from release packages_Manual input required — this section cannot be populated automatically._
| File / Path | What it contributed |
|---|---|
| CLAUDE.md | Sections 1, 4, 5, 6, 10 — product positioning, architecture overview, KB structure, derivative products, ITI shared library references |
| plugin/gd-chatbot.php | Sections 1, 4, 5 — plugin metadata, activation defaults, system prompt, admin menus, AJAX/REST handlers, streaming, DB schema |
| plugin/README.md | Sections 1, 8 — product description, v2 migration notes |
| plugin/readme.txt | Sections 1, 8 — WordPress.org-style readme, changelog entries |
| plugin/CHANGELOG.md | Section 8 — version history through v2.0.6 |
| plugin/includes/class-claude-api.php | Sections 5, 7 — full KB loading, optimized KB loading, guardrails (hardcoded), Bahr Gallery enforcement |
| plugin/includes/class-chat-handler.php | Sections 5, 7 — multi-source context assembly (KB, Pinecone, Tavily, Knowledgebase Loader, AI Power) |
| plugin/includes/class-query-optimizer.php | Sections 5, 7 — intent detection for context selection |
| plugin/includes/class-context-builder.php | Sections 5, 7 — intent-based KB fragment loading |
| plugin/includes/class-response-enricher.php | Sections 2, 7 — song detection in responses, streaming platform links |
| plugin/includes/class-streaming-service-manager.php | Section 7 — OAuth management, multi-platform unified search |
| plugin/public/class-chatbot-public.php | Section 1 — public-facing shortcode, theme loading |
| plugin/context/core/band-and-history.md | Section 5 — example of core knowledge content depth |
| docs/IMPLEMENTATION-SUMMARY.md | Sections 6, 8 — implementation timeline (~30 hours), v2.2.0 completion date |
| docs/ADMIN-GUIDE-STREAMING-SERVICES.md | Section 7 — streaming platform OAuth setup documentation |
| releases/README.md | Sections 1, 8 — v2.2.0 package stats (398KB, 125 files, 31 setlist CSVs, 2,340+ shows) |
| archive/README.md | Section 10 — v1 archive context |
The vibe coding wave has a specific implication for niche community products like GD Chatbot: anyone can now build a “Grateful Dead chatbot” by wrapping ChatGPT with a custom GPT configuration. And someone has — Cosmic Charlie is exactly that, a GPT-based Grateful Dead specialist offering concert archive access, personalized recommendations, and AI-generated psychedelic artwork. The barrier to entry for domain-specific AI chatbots has dropped to nearly zero. What has not dropped is the barrier to building one with genuine depth: 60+ curated knowledge documents, 31 years of structured setlist CSVs covering 2,340+ shows, intent-based context assembly with token budget management, multi-source RAG (Pinecone, Tavily, bundled KB), and a music discovery layer spanning Archive.org plus five commercial streaming services.
LLM convergence affects this product differently than most. Claude, GPT-5, and Gemini 2.5 all know general Grateful Dead facts. The model’s baseline knowledge of the Dead is good enough that a well-prompted generic chatbot can answer common questions adequately. GD Chatbot’s advantage is not that Claude knows more about the Dead in general — it is that the RAG pipeline injects specific, curated, accuracy-checked knowledge (supplementary deep dives, academic papers, gear specifications, gallery guides, disambiguation rules for confusing song titles) and enforces accuracy guardrails that generic models do not have. The Bahr Gallery must always be “Oyster Bay, Long Island, NY” — a guardrail that exists because the knowledge base itself contains conflicting location data that Claude would otherwise synthesize incorrectly. These are the kinds of domain-specific accuracy problems that generic wrappers will hit and not solve.
The more significant competitive shift is JamBot by JamBase — the first commercial AI chatbot targeting jam band fans, backed by JamBase’s live music data API. JamBot is not GD-specific, but it validates the concept of AI-powered music community tools and has a data partnership advantage (real-time concert data) that GD Chatbot does not. The Grateful Dead community’s openness to AI experiences is further evidenced by the ElevenLabs Jerry Garcia AI voice clone endorsed by the Garcia estate.
The competitive landscape for GD Chatbot spans AI chatbots, streaming platforms, setlist databases, and community tools:
| Competitor | Type | AI | Key Strength | Key Weakness |
|---|---|---|---|---|
| JamBot (JamBase) | AI concert chatbot | Yes | Live concert data API, multi-artist scope | Not GD-specific, no deep domain KB, no Archive.org integration |
| Cosmic Charlie (GPT) | GD AI chatbot | Yes | Concert archives, personalized recs, AI artwork | GPT wrapper without custom RAG; no streaming integration; no setlist CSV depth |
| Nugs.net | Streaming platform | No | 30,000+ official recordings, video, multi-device. $12.99/mo | No conversational AI, no knowledge synthesis |
| Relisten | Free streaming | No | 75+ bands, 1M+ recordings, iOS/Android/Mac | No AI, browse-only |
| DEADSHOWZ | Setlist search ($2.99) | No | Multi-criteria search (AND/OR/NOT), DeadBase-curated, offline | No AI, no conversational interface |
| Deadhead Archives | Archive.org browser | No | 100K+ downloads, SBD identification, year-based browsing | No AI, limited search |
| HeadyVersion | Community ratings | No | Fan-voted best versions per song, leaderboard, comments | No AI, no streaming integration |
Eroded differentiators:
Still unique / defensible: Custom RAG with 60+ bundled KB files (vs. Cosmic Charlie’s GPT wrapper), 31-year structured setlist CSV data, intent-based token-optimized context assembly, song detection in AI responses with cross-platform playback links (Archive.org + 5 streaming services), accuracy guardrails hardcoded in PHP (cannot be bypassed via prompt editing), and the proven derivative product pattern (ITI Chatbot, AI News Cafe, Scuba GPT).
The roadmap prioritizes content completeness and the features that transform GD Chatbot from a knowledge tool into a community engagement platform.
Tier 1 — Critical (next build cycle):
Extend setlist data beyond 1995 (L effort) — the single largest content gap. Dead & Company (2015–2024, 200+ shows), Further, The Dead, The Other Ones, and Bobby Weir & Wolf Bros are covered by competitors (DEADSHOWZ, Relisten, Nugs.net, Deadhead Archives) but not by GD Chatbot. Users asking about recent shows hit a hard data wall.
Sync CHANGELOG, readme.txt, and plugin README to v2.2.0 (S effort) — housekeeping prerequisite. The formal CHANGELOG only covers through v2.0.6; the readme.txt stable tag is stuck at 2.0.3; one option key uses the old prefix. This must be resolved before any release.
“Best version” recommendation engine (M effort) — when users ask “What’s the best version of Scarlet Begonias?”, Claude should reference curated version quality data sourced from HeadyVersion community consensus, Deadbase ratings, and KB supplementary content, with explanation of why each version is notable. This addresses the single most common question type in r/gratefuldead.
Personalized show recommendations (M effort) — track user conversation preferences (favorite eras, songs, band members, venues) and use Claude to generate tailored “you should listen to” suggestions. This turns the chatbot from a reference tool into a discovery companion.
Tier 2 — High Value: Guided listening journeys (narrative-driven experiences like “The 1977 Tour” or “Jerry’s Guitar Evolution”), user favorites and listening history, show comparison mode, SBD/AUD recording identification, and conversation analytics dashboard.
Tier 3 — Strategic: Community version ratings (inspired by HeadyVersion’s model but integrated into the conversational experience), AI-generated show posters (responding to Cosmic Charlie’s artwork feature), multi-model support for cost flexibility, tour context cards, Dead & Company knowledge expansion, and proactive conversation follow-up suggestion chips.
Tier 4 — Exploratory: Voice interface (with potential Jerry Garcia AI voice integration), shareable conversation cards for social distribution, Archive.org recording quality auto-assessment, setlist trivia game mode, WordPress.org plugin directory listing, and PWA/offline mode.
The dependency logic: post-1995 setlist data and documentation sync are prerequisites for everything else. Best version recommendations and personalized suggestions build on the existing KB + setlist infrastructure. Guided listening journeys (Tier 2) depend on both the expanded setlist data and the recommendation engine.
Two new Skills from the April 2026 roadmap cycle support GD Chatbot development:
| Skill | What It Provides |
|---|---|
community-engagement-features |
Design and implementation patterns for user ratings, voting systems, leaderboards, and social sharing in WordPress plugins. Anti-spam/anti-gaming measures, aggregation algorithms, and display widgets. Enables the Tier 3 community version ratings feature and social sharing cards. |
guided-content-journeys |
Design patterns for narrative-driven sequential content experiences combining knowledge delivery with action steps (listening, reading, exploring). Journey structure, branching logic, progress tracking, and content curation methodology. Directly enables the Tier 2 guided listening journeys feature. |
The existing grateful-dead-historian Skill (Codex) provides deep domain knowledge for KB content development, and the archive-org-client Skill (Cursor) supports the Archive.org integration that powers the music discovery layer.
Strengths: GD Chatbot’s knowledge-to-listening bridge — where songs mentioned in AI responses automatically become clickable playback links across Archive.org and five streaming services — remains unique and is the product’s most compelling UX innovation. The bundled knowledge base (8 core docs, 21 supplementary deep dives, 3 disambiguation guides, 31 years of setlist CSVs) creates information density that neither Cosmic Charlie’s GPT wrapper nor JamBot’s broad multi-artist approach can match. The accuracy guardrails (hardcoded in PHP, not editable via admin prompts) demonstrate RAG hygiene best practices. The architecture has proven replicable: ITI Chatbot, AI News Cafe, and Scuba GPT are derivatives.
Gaps: Setlist data ending at 1995 is a significant limitation — the Grateful Dead’s successor bands have played for nearly as long as the original band at this point. Documentation is out of sync (CHANGELOG covers through v2.0.6; readme.txt stable tag is 2.0.3; shipping version is 2.2.0). There are no user engagement features (favorites, ratings, history) — the chatbot is purely conversational with no persistent user state. Mobile experience is not optimized for the responsive WordPress shortcode. The context/_archive/ directory ships in release packages.
What we’re watching: JamBot by JamBase represents the first well-funded AI entry into the jam band community space. If JamBase builds GD-specific depth (they have the data partnerships to do it), it could become a serious competitor. The ElevenLabs Jerry Garcia voice clone signals that the Garcia estate and the Dead community are receptive to AI-powered fan experiences — which validates the market but could also set expectations for production quality that a WordPress plugin chatbot does not naturally meet. Nugs.net’s simplified pricing ($12.99/mo with 30,000+ recordings) makes the “listen” step of the knowledge-to-listening pipeline increasingly frictionless, which actually benefits GD Chatbot’s value proposition.
GD Chatbot demonstrates ITI’s approach to domain-specific RAG products: deep curated knowledge, strict accuracy discipline, and workflow integration (knowledge → listening) that creates value beyond what generic AI can offer. The product also serves as the architectural reference implementation for ITI’s WordPress chatbot pattern.