1. Scope
This disclaimer applies to Playground for LM, also called 模型广场 in Simplified Chinese environments. Both names refer to the same app.
2. Nature of the App
The app is a local AI model runtime, management, inference, user-configured API model connection, attachment parsing, local RAG / index-assisted answering, and speech-to-text transcription tool. Its features include local model management, local model loading, local chat, generation settings, Markdown and LaTeX rendering, local recording and media import, local Whisper speech-to-text, transcription history playback, image / document / video attachment parsing, local OCR, local document indexing, RAG Beta routing, API model management, local-network API scanning, runtime logs, and model diagnostics. It sells a local tool and native Apple-platform experience, not third-party model weights, third-party model services, cloud AI services, professional transcription services, or third-party intellectual property content.
3. API Models and Cloud-Service Boundaries
The current version is designed primarily for local model inference and local speech recognition, while also allowing users to actively add third-party or local API models. It does not provide a developer-operated cloud model inference or speech-recognition service and does not actively send user prompts, chats, generated results, recordings, imported audio, imported video, or transcripts to online model services because of local inference or local speech-to-text features.
Only when the user actively adds, selects, and uses a third-party or local API model will related request contents be sent to the configured Base URL or corresponding API provider. Users are responsible for confirming the provider, local-network service, API key, model capability, pricing, privacy policy, terms, and acceptable-use rules.
Local-network API scanning runs only after the user starts it, and may be affected by iOS Local Network permission, being on the same Wi-Fi, computer firewalls, VPNs / virtual adapters, service bind address, port, and API authentication. The app does not guarantee that it can discover or connect to LM Studio, Ollama, llama.cpp, LocalAI, or any other local service.
When using API models, prompts, system prompts, conversation context, generation settings, thinking-control settings, attachment-derived results, RAG / index retrieval snippets, and any images, file contents, or parsed file results selected by the user may be sent to the configured API provider or local-network service. Whether an API provider supports images, attachments, long context, thinking controls, or specific generation parameters depends on that service and is not guaranteed by the app.
The app may access the developer website through an in-app WebView or system browser to display the Privacy Policy, Terms of Use, Third-Party Model Statement, Open Source Licenses, Disclaimer, or other public notices. This access is only for displaying legal documents or public information; it is not cloud model inference or an online AI service.
4. Third-Party Model Sources
The app may provide download entries pointing to ModelScope or other public third-party model sources. Model weights, licenses, training-data rights, capabilities, availability, download stability, content risks, and later changes are declared and managed by the corresponding model publisher or third-party platform.
The App Store release does not use official logos from third-party models or platforms. Visual model marks in the interface are app-owned neutral text badges used only to distinguish model families, sources, or compatible formats, and do not indicate official cooperation, authorization, approval, endorsement, or joint operation.
5. User-Imported Models
Users are responsible for the source, authorization, integrity, security, legality, and suitability of models they import. The app usually performs only basic compatibility checks and does not guarantee model provenance, training-data rights, license validity, content safety, or suitability for a particular purpose.
6. Model Files and Checks
Model files are static model weights or data files for local inference. They are not scripts, plugins, dynamic libraries, extensions, or executable programs. Basic local checks do not confirm legal authorization, training-data legality, content safety, commercial availability, or suitability for any purpose.
7. Model Output
Model output may be wrong, inaccurate, outdated, fabricated, incomplete, biased, misleading, offensive, or unsuitable for particular situations. The app does not guarantee output accuracy, completeness, legality, safety, truthfulness, reliability, suitability, timeliness, continuity, or availability.
Attachment parsing, OCR, image understanding, video recognition, document indexing, RAG Beta retrieval, context expansion, and regeneration may be affected by file format, parsing quality, index coverage, model capability, context window, and device resources. They may miss content, misread materials, cite incompletely, reduce generation quality, or fail. RAG Beta / index-assisted answering only means the app attempts to retrieve relevant local snippets; it does not guarantee 100% accuracy or complete coverage of all sources.
8. Speech-to-Text and Audio Processing
The speech-to-text feature depends on local Whisper models, audio decoding, sample conversion, punctuation restoration, paragraphing, and local post-processing. Transcripts may contain missing words, wrong words, sentence-boundary errors, punctuation errors, speaker confusion, language-detection errors, timestamp drift, unnatural paragraphing, silence misclassification, or complete recognition failure.
Users are responsible for ensuring that any audio, video, image, document, code, spreadsheet, attachment, or transcript they record, import, save, transcribe, play, copy, share, or send to an API service has a lawful source and necessary authorization. Users must comply with applicable rules concerning recording, privacy, personal information, copyright, likeness rights, reputation, trade secrets, workplaces, education, healthcare, and communications. Users must not record, transcribe, save, distribute, or upload another person's conversation, meeting, class, protected work, or sensitive information without authorization.
Transcript text, automatic punctuation, paragraphing, and audio timeline mappings are for general reference only and should not be used as the sole basis for formal minutes, legal evidence, medical records, financial records, identity verification, compliance audits, or other high-risk scenarios.
9. No Professional Advice
The app and generated content do not constitute medical, legal, financial, investment, tax, psychological, educational, safety, engineering, academic, or other professional advice. For important rights, health, safety, property, or legal matters, users should consult qualified professionals.
10. Local Content and User Conduct
The app does not proactively review, filter, monitor, upload, or manually inspect local prompts, chats, generated results, conversation records, recordings, imported media, or transcripts by default. Users must not use that local nature as a reason to engage in illegal, infringing, fraudulent, harassing, dangerous, malicious, privacy-invasive, sexual, violent, hateful, discriminatory, child-inappropriate, or otherwise improper conduct.
11. Intellectual Property and Third-Party Rights
Users are responsible for ensuring that model files, prompts, generated content, audio, video, transcripts, code, text, tables, formulas, and other materials they download, import, run, record, transcribe, copy, save, use, or share do not infringe third-party rights.
12. Third-Party Platforms
Third-party platforms may process network request information according to their own privacy policies and terms. The app does not control their servers, domains, rules, model pages, licenses, download speed, file availability, security, data handling, or later policy changes.
13. Device Resources, Background Running, and Deletion
Model loading, generation, downloading, audio transcription, attachment parsing, OCR, image understanding, document indexing, RAG retrieval, context expansion, API connections, and playback may be affected by device resources, system version, memory pressure, thermal state, battery state, network conditions, background limits, API service status, and Apple platform behavior. The app attempts to support user-initiated background model downloads, background file transcription, and background audio playback where permitted, but iOS / iPadOS may interrupt, pause, delay, or terminate those tasks because of resources, permissions, network, battery, system policy, or user action. Model files, recordings, attachment caches, index files, or transcription history deleted through the app are usually not recoverable.
14. Local Data and Diagnostics
The app stores chat records, system prompts, generation settings, API model configurations, runtime logs, preferences, recording files, imported media files, attachment files, attachment caches, OCR results, index chunks / manifests, transcription history, timeline mappings, and some app state data locally on the user's device. Natural-language log diagnostics, API diagnostics, local-network scan results, and RAG / index status are for reference only and do not constitute a technical-support promise, fault determination, repair recommendation, or legal admission.
15. Open Source, Compatibility, and Changes
The app uses some third-party open-source components under their respective licenses. Features, interface, compatibility, model lists, download entries, diagnostics, and error messages may change because of system versions, devices, App Store requirements, model formats, third-party platform rules, or maintenance plans.
16. Purchases, Pro Entitlement, and Apple Relationship
The app is distributed through the Apple App Store. Future versions may change from a paid download to a free download with a Pro entitlement offered through App Store in-app purchase. The free version may include a limited number of local chat messages. Pro may unlock unlimited local chat, local speech-to-text transcription, and other complete local features, as described in the app and App Store product information.
Users who purchased the paid app through the App Store before the free transition will retain full access automatically where technically feasible and allowed by Apple rules, without needing to purchase Pro again. Development builds, TestFlight builds, or testing environments may show a developer or testing status and may be fully unlocked by default for testing and App Review. Purchases, refunds, billing, order records, Family Sharing, taxes, and Apple Account matters are handled by Apple under its rules. The app is not an official Apple product, and Apple does not endorse or guarantee the app, third-party models, model output, or user conduct.
17. Limitation of Liability and Updates
To the maximum extent permitted by applicable law, the developer is not liable for indirect, incidental, special, profit, data, business-interruption, or third-party-claim losses arising from use of or inability to use the app, local models, third-party models, API models, model output, attachment parsing, OCR, RAG / index-assisted answers, context expansion, speech-to-text, transcripts, audio playback, third-party download links, third-party API services, local-network services, runtime logs, diagnostics, or related content. This disclaimer may be updated as features, laws, App Store requirements, third-party platforms, or operations change.