// Solve
Solve a real problem with AI
Start from who you are and what you need: 24 scenarios for 7 audiences, each with steps, costs, alternatives, ready-made stacks and fixes for common problems.
- scenarios
- 24
- answered problems
- 45
- software stacks
- 44
- audiences
- 7
prices from official pages, with check datesevery answer links to official docs
01By audience
Pick the group closest to you; the most relevant scenarios come first.
Writing, translation, knowledge bases, private local AI
- 01✍️ Everyday work: writing, translation and slidesAll major assistants have free tiers with lower limits and fewer models
- 02🧠 Personal knowledge base: AI that answers from my filesNotebookLM is free to use; Obsidian is free; the local route is all open source
- 03🖥️ Run models on your own machineAll software is free
- 04🎁 Find free AI resources and creditsFree APIs usually have rate and daily limits
- 05🧭 Use AI compliantly in mainland ChinaMajor domestic assistants are basically free
- 06🧱 Build apps and sites without codingLovable, v0, Bolt and Replit all have free credits for prototypes
- 07🎙️ Creators: voice-over, podcasts and AI musicWhisper, CosyVoice, GPT-SoVITS and others are free; Suno has free credits (commercial rights on free plans per its terms)
- 08🗓️ Meeting notes and team workflow automationMost notetakers have free minutes; faster-whisper is free locally
- 09🔁 Switch models: your own key, Chinese or local modelsFree APIs exist; see the Free API page
- 10🎓 Learn AI (roadmaps, tutorials)Our courses and curated tutorials are free
- 11💸 Control AI costs and usage limitsccusage, LiteLLM and Portkey open source are free
Stacks
Common problems
- ChatGPT, Claude or Gemini: should you pay, and for which?
- Get AI to answer only from your own files (NotebookLM)
- Build a private ChatGPT on your own computer (Ollama + Open WebUI)
- Transcribe meeting recordings locally (Whisper / faster-whisper)
- Security checks before launching a vibe-coded app
- Use Claude Code with other models (Chinese models, your own gateway)
AI coding, switching models, MCP, cost and safety
- 01👩💻 Code with AI: choose tools, keep qualityCopilot, Cursor, Gemini CLI and others have free tiers; open-source CLIs (Cline, Aider, OpenCode) can use free APIs; verified students get Copilot free
- 02🔁 Switch models: your own key, Chinese or local modelsFree APIs exist; see the Free API page
- 03💸 Control AI costs and usage limitsccusage, LiteLLM and Portkey open source are free
- 04🔌 Connect tools to agents (MCP, CLI, auth)Official SDKs, Inspector and most MCP servers are free
- 05📄 Build RAG and process documentsParsing, vector stores and eval tools all have open-source versions
- 06🛡️ Let agents run safelygitleaks, gVisor, Docker are free
- 07🖥️ Run models on your own machineAll software is free
- 08📈 Evaluate and monitor AI qualityMain tools are open source
- 09📐 Share rules and skills across agentsRules files and most skills are free
- 10🧪 Choose tools and avoid pitfallsOur software profiles show maintenance and licence info for free
- 11🧭 Use AI compliantly in mainland ChinaMajor domestic assistants are basically free
- 12🎁 Find free AI resources and creditsFree APIs usually have rate and daily limits
- 13🎓 Learn AI (roadmaps, tutorials)Our courses and curated tutorials are free
Stacks
Common problems
- LiteLLM 1.107.0 was compromised: am I affected, and what to do
- Cursor vs Claude Code vs Codex vs Copilot: which to choose
- Build a private ChatGPT on your own computer (Ollama + Open WebUI)
- Are API relays / OpenRouter safe?
- MCP server won’t connect: “Connection closed” / MCP error -32000
- Too many MCP tools filling the context window
Private deployment, data-leak prevention, gateways and audit
- 01🏢 Private LLMs and knowledge bases for the enterpriseOne workstation with 24–48 GB VRAM plus open-source software is enough to validate (see self-hosted compute for models)
- 02🎧 AI customer support and sales assistantsDify, FastGPT, MaxKB and n8n can be self-hosted for free (subject to their licences)
- 03💸 Control AI costs and usage limitsccusage, LiteLLM and Portkey open source are free
- 04📄 Build RAG and process documentsParsing, vector stores and eval tools all have open-source versions
- 05🛡️ Let agents run safelygitleaks, gVisor, Docker are free
- 06🧭 Use AI compliantly in mainland ChinaMajor domestic assistants are basically free
- 07📈 Evaluate and monitor AI qualityMain tools are open source
- 08🧪 Choose tools and avoid pitfallsOur software profiles show maintenance and licence info for free
- 09🗓️ Meeting notes and team workflow automationMost notetakers have free minutes; faster-whisper is free locally
- 10📊 Data analysis and reports with AIWrenAI, DB-GPT, Chat2DB community and Metabase open source can be self-hosted
Stacks
Common problems
- LiteLLM 1.107.0 was compromised: am I affected, and what to do
- Self-host Dify on company servers (Docker Compose)
- Employees paste company data into AI: what to do
- Transcribe meeting recordings locally (Whisper / faster-whisper)
- Are API relays / OpenRouter safe?
- Stop agents from deleting files or running dangerous commands
Literature search, reading papers, data analysis, disclosure
- 01🔬 Research: find, read and review literatureZotero, Semantic Scholar, PDFMathTranslate and NotebookLM are free
- 02📊 Data analysis and reports with AIWrenAI, DB-GPT, Chat2DB community and Metabase open source can be self-hosted
- 03🧠 Personal knowledge base: AI that answers from my filesNotebookLM is free to use; Obsidian is free; the local route is all open source
- 04🖥️ Run models on your own machineAll software is free
- 05📄 Build RAG and process documentsParsing, vector stores and eval tools all have open-source versions
- 06🎓 Learn AI (roadmaps, tutorials)Our courses and curated tutorials are free
Stacks
Common problems
Images, video, voice-over, podcasts, music
- 01🎬 Creators: AI images and videoJimeng, Kling, CapCut and others have free credits; ComfyUI is free locally (needs a GPU)
- 02🎙️ Creators: voice-over, podcasts and AI musicWhisper, CosyVoice, GPT-SoVITS and others are free; Suno has free credits (commercial rights on free plans per its terms)
- 03🧱 Build apps and sites without codingLovable, v0, Bolt and Replit all have free credits for prototypes
Studying with AI, free perks, learning to code
- 01📚 Students: study with AI (the right way)Free tiers of Doubao, Kimi, DeepSeek, Gemini and ChatGPT; NotebookLM is free; students get Copilot free
- 02🎁 Find free AI resources and creditsFree APIs usually have rate and daily limits
- 03🎓 Learn AI (roadmaps, tutorials)Our courses and curated tutorials are free
- 04🧠 Personal knowledge base: AI that answers from my filesNotebookLM is free to use; Obsidian is free; the local route is all open source
- 05👩💻 Code with AI: choose tools, keep qualityCopilot, Cursor, Gemini CLI and others have free tiers; open-source CLIs (Cline, Aider, OpenCode) can use free APIs; verified students get Copilot free
- 06🔬 Research: find, read and review literatureZotero, Semantic Scholar, PDFMathTranslate and NotebookLM are free
- 07✍️ Everyday work: writing, translation and slidesAll major assistants have free tiers with lower limits and fewer models
Stacks
Support & sales, meeting notes, automation, building apps
- 01🎧 AI customer support and sales assistantsDify, FastGPT, MaxKB and n8n can be self-hosted for free (subject to their licences)
- 02🗓️ Meeting notes and team workflow automationMost notetakers have free minutes; faster-whisper is free locally
- 03✍️ Everyday work: writing, translation and slidesAll major assistants have free tiers with lower limits and fewer models
- 04🧱 Build apps and sites without codingLovable, v0, Bolt and Replit all have free credits for prototypes
- 05📊 Data analysis and reports with AIWrenAI, DB-GPT, Chat2DB community and Metabase open source can be self-hosted
- 06🏢 Private LLMs and knowledge bases for the enterpriseOne workstation with 24–48 GB VRAM plus open-source software is enough to validate (see self-hosted compute for models)
- 07💸 Control AI costs and usage limitsccusage, LiteLLM and Portkey open source are free
- 08🎬 Creators: AI images and videoJimeng, Kling, CapCut and others have free credits; ComfyUI is free locally (needs a GPU)
- 09🧪 Choose tools and avoid pitfallsOur software profiles show maintenance and licence info for free
Stacks
Common problems
- ChatGPT, Claude or Gemini: should you pay, and for which?
- Self-host Dify on company servers (Docker Compose)
- Employees paste company data into AI: what to do
- Transcribe meeting recordings locally (Whisper / faster-whisper)
- Generate images locally with ComfyUI
- Labelling AI-generated images and video on Chinese platforms
02By scenario
Each scenario page has: steps · cost tiers · alternatives · when not to use · stacks · problems · related model pages.
03Common problems
All 45 →- 1LiteLLM 1.107.0 was compromised: am I affected, and what to do2026-10-22
- 2AI-generated references that don’t exist2026-10-22
- 3MCP server won’t connect: “Connection closed” / MCP error -320002026-10-10
- 4Remote MCP server OAuth fails2026-10-10
- 5Ollama truncates long documents and the model “forgets” earlier text2026-10-10
- 6llama.cpp prints garbage, mixes up roles or never calls tools2026-10-10
- 7vLLM fails at start-up with CUDA out of memory2026-10-10
- 1Get AI to answer only from your own files (NotebookLM)2026-10-22
- 2Build a private ChatGPT on your own computer (Ollama + Open WebUI)2026-10-22
- 3Self-host Dify on company servers (Docker Compose)2026-10-22
- 4Employees paste company data into AI: what to do2026-10-22
- 5Transcribe meeting recordings locally (Whisper / faster-whisper)2026-10-22
- 6Generate images locally with ComfyUI2026-10-22
- 7Labelling AI-generated images and video on Chinese platforms2026-10-22
- 8How to disclose AI use in a paper2026-10-22
- 9Security checks before launching a vibe-coded app2026-10-22
- 10Free AI tools for students2026-10-22
- 1Cursor vs Claude Code vs Codex vs Copilot: which to choose2026-10-22
- 2ChatGPT, Claude or Gemini: should you pay, and for which?2026-10-22
- 3Are API relays / OpenRouter safe?2026-10-12
- 4What model size can my VRAM / RAM run?2026-10-10
- 5Which PDF/document parser to use for RAG2026-10-10
- 6Giving agents memory across sessions: memory layer or rules file?2026-10-10
- 7Choosing an agent framework: LangGraph, OpenAI Agents SDK, Pydantic AI, CrewAI, Mastra2026-10-10
04Ready-made stacks
All 44 stacks →Proven combinations of tools for common jobs.
- 1Creator / self-media · Topics, copy, covers, short video and scheduling.4
- 2Legal document workflow · Summarize, compare and search contracts/case files (human review required).4
- 3Product manager workflow · Research, PRDs, prototypes, roadmaps and eng collaboration.4
- 4Small team collaboration · Connect docs, tasks, meeting notes and automation.4
- 5Meeting notes automation · Auto-transcribe meetings, summarize and push action items.3
- 6Vibe-coding app builders · Ship deployable apps from natural language.4
- 7Designer workflow · From concept to hi-fi: generate, design systems, hand off to coding agents.4
- 8AI video production · From script to finished video: generate, edit, caption and voice.4
- 9AI music generation · Generate usable songs/scores and separate stems.3
- 10Image & design generation · Generate and refine product shots, marketing visuals and 3D assets.4