Terms of Use
Please read these Terms of Use carefully before accessing or using the Đại Việt (daiviet.dev) platform and its associated public APIs. By accessing any part of the site, you agree to become bound by these terms.
01Acceptance of Terms & Policies
By accessing, browsing, or utilizing the web interface, data feeds, or programmatic API endpoints at daiviet.dev, you confirm that you have read, understood, and agree to be bound by these Terms of Use and our accompanying Privacy Policy & Cookie Governance.
If you do not agree to all terms and conditions set forth herein, you must immediately discontinue access to the website and its APIs.
02Platform Mission & Scope of Service
Đại Việt (daiviet.dev) is an independent, community-driven platform created to aggregate and visualize:
- Frontier foundation model evaluations across 19 distinct metrics spanning 10 benchmark platforms (SWE-bench, BFCL, LiveCodeBench, LMSYS Arena, LiveBench, Artificial Analysis, MTEB, Stanford HELM, Epoch AI, etc.).
- Real-time GitHub star velocity and trending rankings computed over 15-minute scheduled ingestion cycles.
- Compound autonomous agent leaderboards ranking Agent System scaffolding + Foundation Model pairings (SWE-bench Verified/Pro, Open Agent Leaderboard, Scale MCP Atlas, GAIA, WebArena).
- Multi-source AI Model Catalog incorporating OpenRouter, Hugging Face Inference Providers, and LiteLLM data with token cost calculations.
All services and datasets are provided free of charge for research, educational, and developer informational purposes.
03Acceptable Use, API Rate Limiting & Cookies
We provide public REST endpoints (/api/trending, /api/benchmarks, /api/agents, and /api/catalog) for developers to access clean, normalized AI metrics. In using these endpoints, you agree to comply with our fair usage terms:
API requests are governed by an edge reverse-proxy limit of 10 requests per second with a burst allowance of 20 requests, coupled with an application sliding-window cap of 60 requests per minute per client IP. Responses include standard headers: X-RateLimit-Limit, X-RateLimit-Remaining, and X-RateLimit-Reset.
Exceeding the rate limit returns an HTTP 429 Too Many Requests status accompanied by a JSON error payload and a Retry-After header indicating seconds to wait. Clients must back off exponentially and respect the retry interval.
You may not deploy distributed IP rotation proxies, spoof client headers, launch denial-of-service (DDoS) floods, or execute parallel crawler storms designed to circumvent rate limiting. Violating clients are subject to automatic temporary or permanent firewall bans.
By accessing the site, you acknowledge that strictly necessary cookies and local storage tokens (daiviet_consent) are deployed to ensure service security and store your privacy selections. Optional analytics telemetry (Google Analytics 4) is deployed strictly upon affirmative consent through our Cookie Consent banner and is governed by our Privacy Policy.
04Network Access Boundaries & Internal APIs
Our architecture maintains a strict boundary between public and internal-tier APIs:
- Public Endpoints:
/api/trending,/api/benchmarks,/api/agents, and/api/catalogare available for open consumption subject to fair rate limits. - Internal-Only Endpoints: Endpoints under
/api/v1/*(including raw benchmark provenance telemetry) are reserved strictly for internal services within RFC 1918 private subnets (10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16) and local loopback interfaces. - Access Denial: Unauthorized external requests attempting to query internal endpoints will automatically receive an HTTP 403 Forbidden response.
05Model Catalog & Workload Cost Estimation Disclaimers
Token cost calculations, blended USD per 1M token estimates, TTFT latency, and throughput (tokens/sec) are mathematical projections computed from publicly available upstream provider rate cards (OpenRouter, Hugging Face Inference Providers, and LiteLLM). They do not constitute financial advice, formal billing quotes, or commercial warranties.
Provider pricing, volume discounts, context window limits, and throughput fluctuate continuously. Always verify current commercial pricing and terms directly with official model hosts and inference providers prior to production deployment.
06Autonomous Agent Benchmarks Disclaimer
Agent pass rates, cost-per-task metrics, and tool accuracy percentages (such as SWE-bench Verified, Open Agent Leaderboard, and Scale MCP Atlas) reflect standardized, sandboxed laboratory evaluations executed under controlled conditions.
Laboratory agent scores do not guarantee bug-free, safe, or deterministic execution when running autonomous agents (such as Antigravity, Claude Code, OpenHands, Cline, or Cursor) in private repositories or enterprise production environments. Users remain solely responsible for sandboxing, code review, and human-in-the-loop oversight.
07Intellectual Property & Cultural Assets
Cultural Visual Identity: The 14-pointed solar star, Chim Lạc waterbird motifs, and Dong Son bronze drum geometric engravings integrated throughout the platform are inspired by Vietnamese cultural heritage fused with modern microprocessor architecture. These brand assets represent the identity of the Đại Việt project.
Open-Source Codebase: The source code of the web application, background ingestion workers, and microservices is licensed under open-source licenses as indicated in our public repository.
Nominative Fair Use of Trademarks: All third-party trademarks, provider brand names, model identifiers (e.g., Anthropic Claude, OpenAI GPT, Google Gemini, Meta Llama, DeepSeek, Mistral AI), and evaluation harnesses belong to their respective owners. They are referenced solely for factual identification, comparative analysis, and academic transparency under nominative fair use principles.
08Disclaimer of Warranties & Limitation of Liability
Provided "AS IS" and "AS AVAILABLE"
To the fullest extent permitted by applicable law, the maintainers, contributors, and hosting providers of Đại Việt (daiviet.dev) disclaim all warranties, express or implied, including merchantability, fitness for a particular purpose, and non-infringement. We do not warrant that data ingestion will be uninterrupted, that cached responses will immediately reflect upstream changes, or that calculations are completely error-free.
In no event shall the platform maintainers, contributors, or partners be liable for any direct, indirect, incidental, special, consequential, or punitive damages arising from the use of or inability to use this platform, its datasets, its cookie preferences, or its programmatic APIs.
09Governing Law, Modifications & Severability
These Terms of Use shall be governed by and construed in accordance with applicable general law principles for open-source digital community services, without regard to conflicts of law provisions. If any provision of these terms is deemed unlawful, void, or unenforceable, that provision shall be severable from these terms and shall not affect the validity and enforceability of remaining provisions.
We reserve the right to modify these Terms of Use at any time. Changes become effective upon publication with an updated effective date. Continued use of the platform after updates constitutes acceptance of the modified terms.