## User Feedback Analysis — 2025-12-14 (based on Discord 2025-12-11 to 2025-12-13 + GitHub activity Dec 2025) ### Snapshot (what users talked about most) Across the provided feedback, the dominant theme is **“I’m trying to run an agent locally / from scratch, and setup breaks in non-obvious ways.”** In a small but consistent set of reports, **5 of 12 distinct actionable items (~42%)** relate to **plugin/provider configuration and install/update issues** (e.g., `TEXT_LARGE`, OpenAI plugin installation failure, DeepSeek/OpenRouter confusion, `.eliza` directory creation, SQL plugin behavior). --- ## 1) Pain Point Categorization (top recurring friction areas) ### 1) Documentation (High frequency, high severity) **Recurring problems** - **Provider/key setup confusion**: users unclear on where to place keys, which key is required (OpenAI vs OpenRouter vs DeepSeek), and whether keys must be connected to Eliza Cloud. - Example: “Do I need to connect that [OpenAI API key] to elizacloud?” (Discord, 2025-12-13) — unanswered. - **Docs location / discoverability breaks**: “Where did packages/docs/ go?” (GitHub #6122, closed) indicates users rely on repo-local docs paths and get stranded when structure changes. - **Migration guidance scattered**: repeated token migration questions (Discord 2025-12-11/12) and wallet-specific steps (SafePal → Phantom) are being answered ad hoc in chat. **Most affected users** - Newcomers attempting first-run or “agent from scratch” builds. - Users trying non-default model providers (DeepSeek, OpenRouter). --- ### 2) Technical Functionality (High frequency, high severity) **Recurring problems** - **Hard failure when inference plugin not registered** leading to confusing runtime errors: - Example: `TEXT_LARGE` error even with “hi” (Discord, 2025-12-13) traced to missing AI plugin registration (OpenAI). - **Plugin install failures due to outdated packages**: - Example: OpenAI plugin installation failure resolved by “run `elizaos update`” (Discord, 2025-12-13). - **Breaking changes / repo churn impacting builds** (types/tests/logging changes): - Example: monorepo cleanup broke types/tests; PR #6218 fixes TypeScript build errors (Discord 2025-12-11; GitHub completed item #6218). **Most affected users** - Builders installing plugins via CLI and expecting “it just works.” - Contributors pulling latest main and encountering build instability. --- ### 3) Performance (Medium frequency, high severity when it hits) **Recurring problems** - **PGLite degradation under swarm-like workloads**: response times reportedly rising from ~10ms to ~900ms (Discord, 2025-12-11; Odilitime). - **Client bundle size concerns**: large JS payloads (syntax highlighting + mermaid) noted at ~2.6MB minified (Discord, 2025-12-11). **Most affected users** - Power users running multi-agent/swarm environments. - Web client users in bandwidth-constrained contexts. --- ### 4) Integration (Medium frequency, medium-to-high severity) **Recurring problems** - **Twitter agent over-consuming API**: “50 per call” request volume (Discord, 2025-12-12; FenrirFawks). - **Payment / x402 integration ambiguity**: interest in integrating x402 payment capabilities into agents (Discord, 2025-12-12), but no clear “blessed path” documented. - **Model provider interoperability confusion**: DeepSeek usage via OpenRouter vs direct endpoint replacement in OpenAI plugin (Discord 2025-12-11; GitHub #6156 closed). **Most affected users** - Builders integrating agents with external services (X/Twitter, payments, alt model providers). --- ### 5) Community (Medium frequency, high severity in trust/safety contexts) **Recurring problems** - **Scam risk during token migration support**: scammers targeted users asking for migration help; community members had to intervene. - Example: Hexx 🌐 warned a user to block/report scammers (Discord, 2025-12-11). - **Support routing confusion**: users asking migration questions broadly; directed to support tickets channel (Discord, 2025-12-12). **Most affected users** - Token holders and newcomers seeking help in public channels. --- ### 6) UX/UI (Low frequency in this dataset, but ongoing) **Recurring problems** - UI polish issues continue to surface but are being addressed (e.g., markdown spacing fixes; “Dashboard button pointer event” issue #6188 closed; weekly report notes many UI cleanups). - **Expectation gap**: users expect errors to be explained in-product; instead they rely on Discord debugging. --- ## 2) Usage Pattern Analysis (actual vs intended) ### How users are actually using elizaOS - **Local-first agent builds**: users import `AgentRuntime`, assemble plugins manually, and expect minimal boilerplate. - Evidence: GitHub issue #6204 includes “agent from scratch” code and expectation that `.eliza` is auto-created. - **Provider experimentation**: users want OpenAI-compatible but also want DeepSeek/OpenRouter, custom endpoints, and “swap-in” compatibility. - **High-tempo plugin integration**: Twitter agents, RSS feeds, MCP server updates, payments (x402) show users treating elizaOS as an integration hub. - **Speculative/prediction market agents** emerging as a concrete use case (Sapience hackathon; betting mechanics discussion, Discord 2025-12-12). ### Mismatches vs intended usage (implied) - **Inference plugin assumed “core” by users**, but is optional/modular in reality—leading to errors like `TEXT_LARGE` when missing. - Users expect **CLI flows to provision keys and configure providers end-to-end**; current experience still requires manual steps and tribal knowledge. ### Feature requests aligning with real usage - **Stronger first-run wizard / CLI automation** for “create → configure provider → run” (reinforced by PR #6216 and PR #6208 focusing on cloud/provider defaults). - **Clear “provider matrix”** (OpenAI / OpenRouter / DeepSeek direct / Eliza Cloud) with copy-pastable env examples. - **Operational tooling**: rate-limit guidance for Twitter, DB recommendations (PGLite vs Postgres), performance baselines. --- ## 3) Implementation Opportunities (solutions per major pain point) ### Pain Point A: Missing/incorrect AI provider plugin causes confusing runtime errors (`TEXT_LARGE`) **Proposed solutions (prioritized)** 1) **High impact / Low effort**: Add a startup preflight check: if no inference provider registered, fail with a single actionable message: - “No inference plugin configured. Install one of: `@elizaos/plugin-openai`, `@elizaos/cloud`, … See docs link.” - Similar approach: **Next.js** and **Vite** provide explicit “missing config” diagnostics at boot. 2) **High impact / Medium effort**: In CLI `elizaos create`, enforce selecting an AI provider and writing validated env vars; block “run” until present (or offer a guided prompt). - Similar approach: **Supabase CLI** prompts and validates project config before running commands. 3) **Medium impact / Medium effort**: Add “doctor” command: `elizaos doctor` that checks plugin registration, env keys, provider reachability, and prints fixes. --- ### Pain Point B: Plugin installation failures due to outdated packages; update command not obvious **Proposed solutions (prioritized)** 1) **High impact / Low effort**: When plugin install fails, CLI should suggest `elizaos update` automatically (detect version mismatch / known failure patterns). - Similar approach: **Homebrew** suggests `brew update` when taps/formula are stale. 2) **High impact / Medium effort**: Add a compatibility resolver that pins plugin versions to core version (or warns clearly when mismatched). - Similar approach: **pnpm** and **Cargo** surface dependency resolution conflicts explicitly. 3) **Medium impact / Medium effort**: Publish a “Supported versions” table for core + official plugins; link it from errors. --- ### Pain Point C: Docs gaps around keys/providers (OpenAI vs OpenRouter vs DeepSeek; Cloud confusion) **Proposed solutions (prioritized)** 1) **High impact / Low effort**: Add a single canonical doc page: **“AI Providers & Keys”** with: - Provider decision tree - Env var names - Examples: OpenRouter w/ DeepSeek, DeepSeek direct endpoint via OpenAI plugin - “Do I need Eliza Cloud?” FAQ - Similar approach: **LangChain** has provider-specific pages + quickstarts. 2) **High impact / Medium effort**: Embed provider setup snippets into CLI output after `create` (“Next steps: export OPENAI_API_KEY=…”) and link to docs. 3) **Medium impact / Medium effort**: Add a “Copy env template” button/command: `elizaos env init --provider openrouter`. --- ### Pain Point D: Database/performance uncertainty (PGLite degradation; Postgres recommendations not centralized) **Proposed solutions (prioritized)** 1) **High impact / Medium effort**: Publish an “Operational Profiles” doc: - “PGLite: dev/small agents” - “Postgres: swarm/prod” - Include symptoms (e.g., rising latency) and migration steps - Similar approach: **Hasura** and **PocketBase** document dev vs prod DB expectations clearly. 2) **Medium impact / Medium effort**: Add runtime warnings when PGLite latency crosses thresholds (“consider Postgres for this workload”). 3) **Medium impact / Higher effort**: Benchmark suite for common workloads (single agent chat, swarm memory writes, twitter ingestion) and publish results. --- ### Pain Point E: External integration pitfalls (Twitter API overuse; x402 unclear path) **Proposed solutions (prioritized)** 1) **High impact / Medium effort**: Add rate-limit guards + request batching in Twitter plugin, and log a per-action request count. - Similar approach: many SDKs (e.g., **Octokit**) expose rate-limit info and encourage backoff. 2) **Medium impact / Medium effort**: Provide “integration recipes” docs: - “Twitter agent: safe polling intervals + caching” - “Payments/x402: reference agent” (even if experimental) 3) **Medium impact / Low effort**: Add a GitHub issue template for integrations requiring: API quotas, observed request counts, reproduction. --- ### Pain Point F: Community safety during migration support (scams) **Proposed solutions (prioritized)** 1) **High impact / Low effort**: Pin a “No DMs / Official support flow” banner in relevant channels and auto-respond when “migrate” is mentioned with the support-ticket link. - Similar approach: major crypto Discords use keyword-triggered safety bots. 2) **High impact / Medium effort**: Verified “Support” role + allow only ticket-based sensitive help; restrict who can post links in migration threads. 3) **Medium impact / Medium effort**: Public migration status page (including exchange-specific like Bithumb), reducing desperation-driven scam exposure. --- ## 4) Communication Gaps (expectations vs reality) ### Where expectations don’t match reality - **Users assume elizaOS includes an AI model by default**, but modular plugin architecture means “no provider configured” is possible—and currently fails with confusing errors (`TEXT_LARGE`). - **Cloud confusion**: at least one user explicitly asked whether OpenAI keys must be connected to Eliza Cloud; lack of an authoritative answer suggests unclear positioning of Cloud vs local. - **Product announcements without context** (“Kled app release”, “Business Eliza”) generate questions but no follow-through in the same channel, creating uncertainty. ### Recurring questions indicating onboarding gaps - “Why is my simplest message failing?” → missing inference plugin / bad setup. - “Which API key do I need for DeepSeek?” → provider/key mapping unclear. - “How do I migrate from wallet X / exchange Y?” → migration guide fragmentation. - “Where did docs go?” → doc discoverability and stable URLs needed. ### Targeted improvements - Add “First Run: Required Components” section in quickstart (runtime + DB plugin + inference plugin). - Add a “Cloud vs Local” explainer with a comparison table (auth, storage, pricing, where keys live). - Enforce that announcements link to a canonical post (release notes / docs / blog). --- ## 5) Community Engagement Insights ### Power users observed (and their needs) - **sayonara**: repeatedly providing accurate provider/db guidance; needs better docs to point to (reduce repeated Q&A load). - **Stan ⚡ / standujar**: core stabilization and architecture work (types/tests/auth); benefits from tighter release/testing gates and clearer contribution checklists. - **jin / Odilitime**: infra + performance + web patches; need structured performance reports and reproducible benchmarks. - **Hexx 🌐**: frontline support + scam spotting; needs better tooling/policies for safer support. ### Common newcomer needs (onboarding friction) - Clear “how to get from zero → working chat” with minimal decision points. - Provider/key setup that doesn’t require Discord back-and-forth. - Safer, official migration support path surfaced earlier. ### Converting passive users into contributors - Create “Docs bounties” tied to the top repeated Discord questions (providers/keys, migration, DB choice). - Add “good first issue” labels for CLI diagnostics, doc improvements, and integration recipes. - Encourage helpers (e.g., sayonara/Hexx) to convert answers into PRs via a lightweight “Answer → Doc PR” workflow. --- ## 6) Feedback Collection Improvements ### Effectiveness of current channels - **Discord** is effective for rapid help but produces **untracked, repeat questions** and **unanswered** items (e.g., OpenAI key vs Cloud, Kled release, redeem). - **GitHub issues** capture reproducible technical problems well (e.g., `.eliza` auto-create expectation #6204) but are underused for onboarding confusion and provider/key questions (many resolved informally in chat). ### Improvements for more structured, actionable feedback - Add a Discord bot command: `/report-bug` that opens a prefilled GitHub issue (includes versions, OS, provider, logs). - Weekly “Top 5 Discord issues” auto-summary posted to GitHub Discussions for triage/closure. - Standardize tags in Discord help threads: `provider`, `cli`, `db`, `migration`, `twitter`, `cloud`. ### Underrepresented segments - **Non-English users** (notably Korean users affected by Bithumb migration) appear in feedback but lack a dedicated, structured update channel. - **Non-crypto builders**: much of the visible community talk is token/migration oriented; general SaaS/enterprise users (implied by “Business Eliza”) are not providing detailed feedback yet. --- ## Prioritized High-Impact Actions (next 2–4 weeks) 1) **Add inference-provider preflight + better fatal errors** (eliminate confusing `TEXT_LARGE` path; link to provider setup docs). 2) **Publish a single canonical “AI Providers & Keys” guide** (OpenAI vs OpenRouter/DeepSeek vs Cloud; include env templates + FAQs). 3) **Ship `elizaos doctor` (or equivalent) + auto-suggest `elizaos update` on known install failures** to reduce Discord-dependent debugging. 4) **Create a migration safety + status hub** (Bithumb/Korean updates + anti-scam warnings + official support routing). 5) **Address high-cost integrations first: Twitter request amplification** (rate limiting/batching + visible request counts in logs).