Menu
← All productsNetwork-specific intelligence layer

nLLM

Intelligence that speaks
infrastructure.

nLLM™ is Invences’ network-specific intelligence layer, designed to connect telemetry, topology, runbooks and incident history to investigation and recommendations. It supports iFabric’s agentic workflows and can connect to configured ITSM processes.

Explore nLLM See architecture
Network-specific intelligence layer

nLLM / How it works

From infrastructure context to guidance.

Prepared knowledge. Current evidence. Grounded reasoning.

Configured corpus, model & operational inputs

nLLM knowledge and reasoning architecturePrepare operational knowledge as a searchable index. At runtime, combine relevant retrieved material with a question and current infrastructure evidence to investigate likely causes and recommend next steps. Knowledge preparation and runtime inference are separate processes. Recommendations enter configured iFabric or ITSM workflows. Consequential actions require the deployment’s authorization and customer or partner execution tools. Knowledge preparation: sources, curated corpus, embeddings and searchable index. Runtime: question and current evidence, retrieval from prepared knowledge, grounded reasoning, recommendation, configured workflow handoff. Current telemetry does not animate into a retrained model. An illustrative insufficient-context path ends in operator review. Animation repeats the explanation, not an operational action.Knowledge preparationScoped sources → searchable contextRuntime reasoningQuestion + current evidencePrepared contextInsufficient contextIllustrative review pathActions require configured authorizationIllustrative sequence · RepeatsRuntime retrieval reads prepared knowledge. This is a reference connection, not model training or a preparation cycle per request.sources to curatecurate to indexquestion to retrieveretrieve to reasonreason to recommendrecommend to handoffIllustrative proposed review when context is insufficient. No workflow handoff or execution.Select relevant runbooks, standard operating procedures, vendor documentation and incident history within the agreed knowledge scope.SourcesRunbooks · docs · historyPrepare the selected material for the configured corpus. Knowledge ownership, permitted content and the preparation process are defined for the deployment.CurateScoped corpusCreate or update the searchable knowledge index through the configured preparation pipeline. This prepares retrieval context; it does not depict model training or an instant model update.IndexEmbeddings + lookupReceive the operational question with relevant current telemetry, topology and configuration evidence from configured sources. These inputs belong to this runtime request.QuestionCurrent evidenceLook up relevant material from the prepared knowledge index for the current question. Runtime evidence does not flow into an automatic corpus or model update.RetrieveRelevant knowledgeUse the configured infrastructure reasoning model with current evidence and retrieved operational knowledge to investigate likely causes. The quality and coverage of the available context constrain the result.ReasonGrounded contextProduce infrastructure guidance or a proposed next step grounded in the available evidence and procedures. A recommendation is not permission to execute it.RecommendProposed next stepPass guidance to the configured iFabric or ITSM workflow. The deployment’s authorization system and customer or partner execution tools govern any consequential action; execution is outside this reasoning diagram.HandoffAuthorization nextIllustrative proposed operator review: gather or check missing context. No workflow handoff or execution. Confirm exception handling in deployment scope.Review contextOperator next stepnLLM knowledge and reasoning architecturePrepare operational knowledge as a searchable index. At runtime, combine relevant retrieved material with a question and current infrastructure evidence to investigate likely causes and recommend next steps. Knowledge preparation and runtime inference are separate processes. Recommendations enter configured iFabric or ITSM workflows. Consequential actions require the deployment’s authorization and customer or partner execution tools. Knowledge preparation: sources, curated corpus, embeddings and searchable index. Runtime: question and current evidence, retrieval from prepared knowledge, grounded reasoning, recommendation, configured workflow handoff. Current telemetry does not animate into a retrained model. An illustrative insufficient-context path ends in operator review. Animation repeats the explanation, not an operational action.Knowledge preparationScoped sources → searchable contextRuntime reasoningQuestion + current evidencePrepared contextInsufficient contextIllustrative review pathIllustrative sequence · RepeatsRuntime retrieval reads prepared knowledge. This is a reference connection, not model training or a preparation cycle per request.sources to curatecurate to indexquestion to retrieveretrieve to reasonreason to recommendrecommend to handoffIllustrative proposed review when context is insufficient. No workflow handoff or execution.Select relevant runbooks, standard operating procedures, vendor documentation and incident history within the agreed knowledge scope.SourcesRunbooks · docs · historyPrepare the selected material for the configured corpus. Knowledge ownership, permitted content and the preparation process are defined for the deployment.CurateScoped corpusCreate or update the searchable knowledge index through the configured preparation pipeline. This prepares retrieval context; it does not depict model training or an instant model update.IndexEmbeddings + lookupReceive the operational question with relevant current telemetry, topology and configuration evidence from configured sources. These inputs belong to this runtime request.QuestionCurrent evidenceLook up relevant material from the prepared knowledge index for the current question. Runtime evidence does not flow into an automatic corpus or model update.RetrieveRelevant knowledgeUse the configured infrastructure reasoning model with current evidence and retrieved operational knowledge to investigate likely causes. The quality and coverage of the available context constrain the result.ReasonGrounded contextProduce infrastructure guidance or a proposed next step grounded in the available evidence and procedures. A recommendation is not permission to execute it.RecommendProposed next stepPass guidance to the configured iFabric or ITSM workflow. The deployment’s authorization system and customer or partner execution tools govern any consequential action; execution is outside this reasoning diagram.HandoffAuthorization nextIllustrative proposed operator review: gather or check missing context. No workflow handoff or execution. Confirm exception handling in deployment scope.Review contextOperator next step

Retrieve context. Reason. Hand off guidance.

Knowledge / evidenceReasoning / guidanceReview

Recommendations enter configured iFabric or ITSM workflows. Consequential actions require the deployment’s authorization and customer or partner execution tools.

Prepare operational knowledge as a searchable index. At runtime, combine relevant retrieved material with a question and current infrastructure evidence to investigate likely causes and recommend next steps. Knowledge preparation and runtime inference are separate processes.

Architecture & operating boundaries

Prepare operational knowledge as a searchable index. At runtime, combine relevant retrieved material with a question and current infrastructure evidence to investigate likely causes and recommend next steps. Knowledge preparation and runtime inference are separate processes.

  1. Sources. Select relevant runbooks, standard operating procedures, vendor documentation and incident history within the agreed knowledge scope.
  2. Curate. Prepare the selected material for the configured corpus. Knowledge ownership, permitted content and the preparation process are defined for the deployment.
  3. Index. Create or update the searchable knowledge index through the configured preparation pipeline. This prepares retrieval context; it does not depict model training or an instant model update.
  4. Question. Receive the operational question with relevant current telemetry, topology and configuration evidence from configured sources. These inputs belong to this runtime request.
  5. Retrieve. Look up relevant material from the prepared knowledge index for the current question. Runtime evidence does not flow into an automatic corpus or model update.
  6. Reason. Use the configured infrastructure reasoning model with current evidence and retrieved operational knowledge to investigate likely causes. The quality and coverage of the available context constrain the result.
  7. Recommend. Produce infrastructure guidance or a proposed next step grounded in the available evidence and procedures. A recommendation is not permission to execute it.
  8. Handoff. Pass guidance to the configured iFabric or ITSM workflow. The deployment’s authorization system and customer or partner execution tools govern any consequential action; execution is outside this reasoning diagram.

This is an illustrative architecture, not a live product demonstration or customer result. The knowledge-preparation band and runtime band are separate: current telemetry supplies request context, not instant model retraining. The prepared-index connection represents retrieval access rather than a fresh preparation cycle for every question. Define corpus ownership, model configuration, evidence sources, access rules and workflow integrations for the environment. No particular model version, corpus size or reasoning accuracy is implied. Review for insufficient context is an illustrative proposed exception flow to confirm in deployment scope. Animation timing is illustrative; repetition restarts the explanation, not an operational action.

Views

Live reasoning. A question and current infrastructure evidence enter runtime retrieval. Relevant prepared knowledge grounds the reasoning and resulting recommendation. Guidance reaches a configured workflow, where any consequential action still requires authorization. This path does not retrain the model or update the knowledge corpus.

Knowledge preparation. Selected runbooks, documentation and incident history are prepared as a scoped corpus and searchable index through the configured knowledge pipeline. This process is separate from a live request and ends before runtime inference; it does not depict a model-training cycle.

Insufficient context. The available runtime evidence and retrieved context are insufficient to support a next step. This illustrative review path ends with an operator gathering or checking context. It does not continue to workflow handoff or execution. Confirm this exception handling in the deployment scope; it is not a claim of confirmed native behavior.

nLLM / In focus

Operational knowledge.
Informed action.

nLLM supplies reasoning for iFabric’s autonomy capabilities and can connect to an enterprise’s existing ITSM workflows. Its guidance draws on the systems, evidence and procedures relevant to the environment.

Capabilities

01

Connect operational knowledge

Bring telemetry, topology, incident history, runbooks and vendor documentation into the reasoning process.

02

Diagnose and recommend

Use infrastructure context to investigate incidents, explain likely causes and propose next steps.

03

Support governed action

Connect reasoning to iFabric and ITSM workflows, with actions subject to the deployment’s policy and authorization.

Start with your environment.

Talk to our team
Our platform directionExplore the platforms iFabric · iCore · Powered by nLLM