## User Feedback Analysis — 2026-05-10 (based on May 7–9 inputs) ### Data notes & sample size - **Sources in scope:** Discord (#discussion, #coders) + recent GitHub issues/PR review notes included in the aggregate. - **Distinct feedback threads identified:** **10** (used as denominator for percentages below). *Caveat:* Discord traffic was low-volume; severity weighting leans heavily on GitHub issues because they include reproducible failures. --- ## 1) Pain Point Categorization (top recurring 5–7) ### 1) Integration — **Connectors/auth requirements are shifting and unclear** (3/10 threads, **30%**) **Examples** - **X/Twitter integration now requires X API**; users asked if older login methods still work and expressed cost concerns (Discord: *guru0 ↔ odilitime*, May 7). - **Telegram delivery reliability issues** surfaced in GitHub: dual pollers consuming the same bot updates causing **silent message loss** (GitHub issue #7245). - **Plugin availability/packaging problems**: Hyperfy plugin removed (404) due to core version mismatch; users resorted to a DM’d zip (Discord: *binarycookies ↔ odilitime*, May 7). **Impact** - Breaks “it should just connect” expectations; also creates **hidden costs** (X API) and **hidden unreliability** (Telegram message drops). --- ### 2) Technical Functionality — **Fresh installs can be broken by missing/misaligned runtime artifacts** (4/10 threads, **40%**) **Examples** - `dev-ui.mjs` references a stealth preload file that doesn’t exist on fresh clone → Anthropic OAuth subscription calls fail with **401** (GitHub issue #7210). - `plugin-sql` runtime-migrator missing drizzle table defs → fresh PGLite boot never creates key tables; chat becomes effectively unusable (GitHub issue #7222). - Duplicate `MiladyClient` class → prototype augmenters attach to the wrong class; many client methods undefined; UI freezes (GitHub issue #7244). - Codex CLI auth detection misses modern `tokens.access_token` format → UI shows “install required” incorrectly; provider auto-enable fails (GitHub issue #7243). **Impact** - These are **onboarding killers**: “fresh clone → dev → chat” fails in multiple paths, often with misleading symptoms. --- ### 3) Documentation — **Status pages + “what changed” guidance are failing** (2/10 threads, **20%**) **Examples** - elizaos.github.io `summary/day` page stuck since May 4 (Discord: *stan0473*, May 7). - X API requirement change needs explicit docs update (Discord: *guru0*, May 7). **Impact** - Users rely on summaries and docs as “source of truth”; when stale, Discord becomes the de facto documentation. --- ### 4) Community / Support Process — **Help requests go unanswered; security flags lack actionable detail** (2/10 threads, **20%**) **Examples** - A concrete dev question (“movement functionality” in a chat/bubble UI project) received **no responses** (Discord: *binarycookies*, May 9). - A one-word “scam” warning posted with no context, leaving others unable to act (Discord: *dieantwoord1337*, May 9). **Impact** - Newcomers churn; security posture is weakened by low-signal alerts. --- ### 5) UX/UI — **When underlying APIs fail, UX fails “silently” or misleadingly** (2/10 threads, **20%**) *(This overlaps with technical issues but is worth isolating because it’s user-facing behavior.)* **Examples** - Telegram dual poller issue manifests as “bot sometimes ignores me” with no clear UI warning (GitHub #7245). - Codex subscription mis-detected as “install required” even when authenticated (GitHub #7243). **Impact** - Users misattribute failures to themselves; higher support burden and lower trust. --- ## 2) Usage Pattern Analysis (actual vs intended) ### Observed real-world usage patterns 1. **Connector-first adoption**: Users are trying to get value by connecting platforms (X, Telegram, Discord) early; connector breakage therefore blocks perceived “core” value. 2. **Ecosystem building via plugins/starters**: The Hyperfy incident shows users treat starter plugins as stable entry points; when a repo disappears (404), they fall back to informal distribution (DM zip). 3. **Full-stack production builds around elizaOS** (Discord May 8 showcase): modern web stacks (Next.js/React/TS) + multi-agent frameworks (LangGraph/CrewAI) + RAG (Pinecone/Weaviate) indicate users combine elizaOS with a broader agent ecosystem rather than using it alone. 4. **Fork-driven workflows (milady references in GitHub issues)**: Users may experience elizaOS primarily through downstream wrappers/templates; detection logic and “fresh clone” assumptions must match that reality. ### Emerging / unexpected use cases - **3D/social world agent integrations** (Hyperfy plugin interest). - **Token/market narrative discussions** (ElizaOS coin analysis) occurring alongside dev support—suggests mixed audience needs in the same channels. ### Feature requests implied by usage - A **versioned plugin compatibility matrix** (core ↔ plugin versions) and automated install guidance. - A **connector health dashboard** (polling status, message ingestion, auth state) to make “silent failures” visible. --- ## 3) Implementation Opportunities (solutions per major pain point) ### A) Plugin removal / version mismatch (Hyperfy 404 + DM zip) **Solutions** 1. **Restore via “deprecated” archive + compatibility badge** (High impact / Low–Med difficulty) - Keep the repo available read-only; mark as incompatible with latest core; provide last-known compatible core version. 2. **Introduce a plugin registry with semver + compatibility constraints** (High impact / Med difficulty) - Similar to **Home Assistant HACS** or **VS Code Marketplace** patterns: plugin declares `engines.elizaos: ">=2.0.0-alpha.530 <2.0.0-alpha.560"`. 3. **Automated CI to detect plugin breakage against core nightly** (High impact / Med difficulty) - Run minimal smoke tests per plugin (build + load + one action call). --- ### B) Fresh-clone runtime breakages (missing artifacts, schema drift, duplicate classes) **Solutions** 1. **“Fresh clone” CI job that runs the golden path** (High impact / Med difficulty) - Steps: clone → install → run dev → open UI route → send one chat message → assert no fatal errors. - Similar to how **Next.js** and **Supabase** maintain “create-*” template smoke tests. 2. **Fail-loud guards instead of silent fallbacks** (High impact / Low difficulty) - Example: if `stealth.claude === true` and preload is missing, **log a warning/error** (GitHub #7210 suggested this). 3. **Single source of truth checks for schemas** (High impact / Med difficulty) - Add a boot-time assertion that every abstract schema exported has a corresponding migrator/drizzle table, or generate drizzle tables from one canonical definition (GitHub #7222 root cause). --- ### C) Connector reliability & “silent message loss” (Telegram dual pollers) **Solutions** 1. **Enforce single ownership per connector token** (High impact / Low–Med difficulty) - If wrapper polling is enabled, auto-disable upstream plugin poller (or vice versa) and surface a UI notice. 2. **Add connector-level health telemetry** (Med–High impact / Med difficulty) - Track last update received, poller instance id, and dropped update count; expose in Settings. 3. **Message ingestion invariants + warnings** (High impact / Med difficulty) - If multiple pollers are detected, warn loudly. Comparable to how some **Kubernetes controllers** detect “multiple leaders” and emit events. --- ### D) Documentation/site reliability (summary page stuck; X API requirement change) **Solutions** 1. **Docs “Breaking/Behavior changes” changelog per connector** (High impact / Low difficulty) - Put “X API now required” in a pinned doc + release notes; link from connector settings UI. 2. **Automated monitoring for docs freshness** (Med impact / Low difficulty) - If daily summary hasn’t updated in >24h, open a GitHub issue automatically. 3. **In-product notices** (Med impact / Med difficulty) - Show “X posting requires paid API tier” inline, including rough cost expectations and alternatives (read-only mode, manual export). --- ### E) Support responsiveness + security reporting quality **Solutions** 1. **Help-request triage bot / workflow** (High impact / Low–Med difficulty) - When a question is posted in #coders, bot asks for: repo link, minimal repro, logs; adds a thread + tags helpers. 2. **Dedicated “Security Reports” intake template** (Med impact / Low difficulty) - Convert one-word “scam” into actionable: what message, who, link, screenshot, recommended action. 3. **Office hours / pairing queue** (Med impact / Med difficulty) - Especially for newcomers like *binarycookies*; increases retention and contributor conversion. --- ## 4) Communication Gaps (expectations vs reality) ### Recurring expectation mismatches - **“Login/auth should work without paid APIs”** vs current reality for X: users expected prior login paths; now **X API is required** (Discord May 7). - **“Official plugins are stable and discoverable”** vs Hyperfy plugin removal: users experienced a sudden 404 and relied on private distribution (Discord May 7). - **“If something is wrong, the UI will tell me”** vs silent failures: - Telegram messages dropped with no clear indicator (GitHub #7245). - Codex login present but UI says “install required” (GitHub #7243). ### Onboarding/documentation gaps indicated by questions - “Do I need X API or does login still work?” → docs + connector UI need a single definitive answer. - “Where did the Hyperfy plugin go?” → need an official plugin lifecycle policy: maintained / experimental / deprecated. ### Specific improvements - Add a **Connector Requirements** table (cost, auth type, rate limits, required tiers) and link it directly in Settings. - Add a **Plugin Lifecycle & Compatibility** page with deprecation policy and recovery steps. --- ## 5) Community Engagement Insights ### Power users & their needs (observed) - **odilitime**: acted as primary responder; provided authoritative integration info and workarounds (X API requirement; Hyperfy zip). Needs tooling to avoid being a single point of failure (knowledge base, triage automation). - **Sw4pIO (GitHub)**: high-signal bug reporter with detailed repro and proposed fixes (#7210, #7222, #7243–#7245). Needs clearer “where to report what” and faster maintainer acknowledgment loops. - **harry346165 / radu83._49711 / itssowenn4462**: builders showcasing stacks and looking to collaborate—good candidates for “community solutions” writeups/tutorials. ### Newcomer friction signals - *binarycookies* asked for help and received none (May 9), and also struggled with plugin discovery/compatibility (May 7). This is a classic “early churn” risk. ### Converting passive users into contributors - Create “**Good First Triage**” roles: turn unanswered Discord questions into GitHub Discussions with a template. - Offer “**Plugin Maintainer**” badges/ownership for ecosystem plugins (e.g., Hyperfy), including CI support and compatibility gates. --- ## 6) Feedback Collection Improvements ### Current channel effectiveness - **Discord**: fast, but low follow-through; questions go unanswered; solutions are sometimes moved to DMs (Hyperfy zip), reducing communal knowledge. - **GitHub issues**: high-quality, reproducible, actionable; currently the best signal for onboarding breakages. ### Improvements for more structured, actionable feedback 1. **Discord → GitHub bridge**: a bot command to file a GitHub issue from a Discord message (pre-filling repro fields). 2. **Standardized templates**: - “Connector bug” (token type, polling mode, expected/actual, logs) - “Fresh install broken” (OS, bun/node, steps, first error line) 3. **Tagging & routing**: label issues by subsystem (vault, plugin-sql, telegram, x) and auto-request reviews from code owners. ### Underrepresented segments - **Non-Discord users** (self-hosters, enterprise) likely under-report; add lightweight web intake or GitHub Discussions prompts. - **Users blocked by cost constraints** (X API) need a clear place to request alternatives or read-only modes. --- ## Prioritized High-Impact Actions (next 2–4 weeks) 1. **Establish plugin lifecycle + restore availability for removed starters (e.g., Hyperfy) with compatibility metadata** *Outcome:* eliminates 404 surprises; reduces DM-only distribution; improves trust. 2. **Add a “fresh clone golden path” CI smoke test and fail-loud guards for missing artifacts** *Outcome:* prevents regressions like missing stealth preload, schema drift, duplicate client classes from reaching users. 3. **Connector reliability hardening: enforce single poller ownership for Telegram + add basic health indicators in Settings** *Outcome:* removes silent message loss; reduces “it sometimes ignores me” reports. 4. **Update connector documentation + in-UI requirements messaging for X (API required, cost implications, alternatives)** *Outcome:* aligns expectations; reduces repeated questions and frustration. 5. **Implement support triage: help-request template + bot-driven escalation for unanswered #coders posts** *Outcome:* improves newcomer retention and turns questions into searchable, reusable answers.