{ "manifest_version": 1, "name": "nr-llm", "slug": "/t3x-nr-llm/", "main_version": "0.41.5", "latest_release": "v0.41.5", "release_date": "2026-10-25", "docs_version": "0.41.5", "php_versions": [ "8.2" ], "typo3_versions": [ "13.4", "14.3" ], "stage": "beta", "owner": "Netresearch DTT GmbH — nr-llm maintainers", "last_verified": "2026-08-11", "license": "GPL-2.0-or-later", "repository": "https://github.com/netresearch/t3x-nr-llm", "documentation": "https://github.com/netresearch/t3x-nr-llm/blob/main/Documentation/Index.rst", "demo": null, "support": "https://www.netresearch.de/kontakt/", "providers": [ "OpenAI", "Anthropic Claude", "Google Gemini", "Ollama (local)", "OpenRouter", "Mistral", "Groq", "Azure OpenAI", "any OpenAI-compatible endpoint (vLLM, LocalAI, LiteLLM)" ], "capabilities": [ { "id": "provider-abstraction", "label": "One provider abstraction for every extension", "detail": "Chat, completion, streaming, tool calling, translation, vision and embeddings behind injectable services.", "state": "implemented", "evidence": "https://github.com/netresearch/t3x-nr-llm/blob/main/Documentation/Developer/Index.rst" }, { "id": "provider-fallback", "label": "Fallback chains per configuration", "detail": "Retried on connection errors, HTTP 5xx and rate limits, in the order the configuration declares.", "state": "implemented", "evidence": "https://github.com/netresearch/t3x-nr-llm/blob/main/Classes/Provider/Middleware/FallbackMiddleware.php" }, { "id": "rag", "label": "Site-content retrieval with citations", "detail": "Answers reference the pages they came from rather than asserting unsourced text.", "state": "implemented", "evidence": "https://github.com/netresearch/t3x-nr-llm/blob/main/Documentation/Administration/Tools.rst" }, { "id": "tools", "label": "Built-in function-calling tools in toggleable groups", "detail": "Read-only by default; the writing tools ship disabled.", "state": "implemented", "evidence": "https://github.com/netresearch/t3x-nr-llm/blob/main/Documentation/Administration/Tools.rst" } ], "security_controls": [ { "id": "vault-keys", "label": "API keys stored as vault identifiers", "detail": "Envelope encryption via nr-vault; nr-llm never stores or logs a raw key.", "state": "implemented", "evidence": "https://github.com/netresearch/t3x-nr-llm/blob/main/Documentation/Adr/Adr012ApiKeyEncryption.rst" }, { "id": "guardrails", "label": "Guardrail pipeline across input, output, reasoning and streamed chunks", "detail": "Screens and can redact; secret patterns are denied on the egress path.", "state": "implemented", "evidence": "https://github.com/netresearch/t3x-nr-llm/blob/main/Documentation/Adr/Index.rst" }, { "id": "hitl", "label": "Human-in-the-loop approval for sensitive actions", "detail": "A run that reaches an approval-gated tool suspends until a person decides.", "state": "implemented", "evidence": "https://github.com/netresearch/t3x-nr-llm/blob/main/Documentation/Adr/Index.rst" }, { "id": "permissions", "label": "Per-capability backend-group permissions", "detail": "Who may use which capability is a backend group setting, not a code constant.", "state": "implemented", "evidence": "https://github.com/netresearch/t3x-nr-llm/blob/main/Documentation/Administration/Index.rst" }, { "id": "supply-chain", "label": "Signed releases with provenance and SBOMs", "detail": "Cosign signature, SHA-256 checksums, SPDX and CycloneDX SBOMs, SLSA build provenance.", "state": "implemented", "evidence": "https://github.com/netresearch/t3x-nr-llm/releases" } ], "cost_controls": [ { "id": "user-budgets", "label": "Per-user budgets", "detail": "Requests, tokens or estimated cost, daily or monthly, per backend user.", "state": "implemented", "evidence": "https://github.com/netresearch/t3x-nr-llm/blob/main/Documentation/Administration/UserBudgets.rst" }, { "id": "analytics", "label": "Usage and cost analytics", "detail": "Estimated cost and usage per provider, model, service and user.", "state": "implemented", "evidence": "https://github.com/netresearch/t3x-nr-llm/blob/main/Documentation/Administration/Analytics.rst" }, { "id": "cheaper-fallback", "label": "Fallback to a cheaper or local model", "detail": "A configuration can name fallbacks that cost less, including a self-hosted endpoint.", "state": "implemented", "evidence": "https://github.com/netresearch/t3x-nr-llm/blob/main/Classes/Domain/DTO/FallbackChain.php" } ], "evidence": [ { "type": "code", "label": "Source", "url": "https://github.com/netresearch/t3x-nr-llm" }, { "type": "adr", "label": "Architecture Decision Records", "url": "https://netresearch.github.io/t3x-nr-llm/adr/" }, { "type": "docs", "label": "Documentation", "url": "https://github.com/netresearch/t3x-nr-llm/blob/main/Documentation/Index.rst" }, { "type": "release", "label": "Releases with SBOM and provenance", "url": "https://github.com/netresearch/t3x-nr-llm/releases" } ], "ai": { "intended_purpose": "Shared, administrator-controlled access to language models for TYPO3 extensions: one provider configuration, one set of permissions and budgets, reused by every extension on the site.", "excluded_uses": [ "Unattended decisions about people — hiring, credit, benefits, access or any other consequential outcome.", "Autonomous content changes without a human approval.", "A source of facts. Model output is untrusted content and is treated as such.", "A compliance control in its own right. It supports controls; it does not make an organisation conformant." ], "models": [ "Whatever the administrator configures: a commercial API, an EU-hosted provider, an OpenAI-compatible endpoint, or a self-hosted model." ], "data": [ "The prompt an extension sends, plus any site content a retrieval tool includes.", "Usage metadata: provider, model, service, backend user, token counts and estimated cost.", "No API keys — those stay in the vault as identifiers." ], "processing_location": [ "own-server", "external-provider" ], "human_oversight": "Writing tools ship disabled and, once enabled, suspend the run for an explicit approval. Everything else is read-only by default.", "permissions": "Per-capability backend-group permissions; the backend module is restricted to administrators.", "logging": "Requests are recorded with provider, model, service, user, token counts and estimated cost. Prompt content is not part of the usage record.", "retention": "Usage records live in the site's database and follow the site's own retention rules. nr-llm sets no expiry of its own.", "cost_control": "Per-user budgets on requests, tokens or estimated cost, daily or monthly, with fallback to cheaper or local models.", "security_controls": [ "API keys as vault identifiers, never plaintext", "Guardrail pipeline across input, output, reasoning and streamed chunks", "Secret-pattern denial on the egress path", "Approval-gated writing tools", "Signed releases with SBOMs and SLSA provenance" ], "known_limitations": [ "Models produce wrong answers with confidence. Nothing here prevents that; the controls bound what a model may do, not what it may say.", "Prompt injection through retrieved site content is possible. Tool permissions and the approval gate are the mitigation, not immunity.", "Cost figures are estimates derived from the model's published pricing, not from the provider's invoice.", "Streaming responses bypass parts of the output pipeline; the trade-off is documented in the ADRs.", "Provider availability, model behaviour and pricing are outside this extension's control." ] } }