> ## Documentation Index > Fetch the complete documentation index at: https://docs.bijection.com/llms.txt > Use this file to discover all available pages before exploring further. # Packaged AI capabilities > Compose models, agents, knowledge, document review and evaluations through existing Bijection components. Bijection packages common AI workflows as components and typed adapters. Your application declares the model, tools, source population and review rules. Native requests, permissions, files and external calls retain their existing authority. Inference, parsing, training and other compute run on external sources. Bijection does not provide a managed compute platform. Components connect those sources to the application's governed data and operations. ## Choose the owner | Capability | Package and entry point | What it owns | | - | - | - | | Models and metering | `@bijectionhq/language-models` | Declared providers, exact model IDs, capabilities, budgets and attempt accounting | | Platform gateway access | `@bijectionhq/language-models/gateway` | Catalog adapter with a fresh scoped token for the exact configured gateway URL | | Typed AI tasks | `@bijectionhq/language-models/tasks` | Validated input/output, bounded context and explicit terminal outcomes | | Managed agents | `@bijectionhq/agents/managed` | Bounded model/tool loops over existing thread storage | | Knowledge collections | `@bijectionhq/rag/collections` | Deterministic chunks, isolated index generations and current source citations | | Document intelligence | `@bijection/documents` | Native source versions, work attempts, semantic evidence, corrections and review | | Durable orchestration | `@bijectionhq/workflow` | Existing durable step continuations around external work | | Workspace assistant | `@bijection/assistants` | Member-owned conversations, authorized tools, source evidence and operation drafts | | AI evaluations | `@bijectionhq/experiments/ai` | Versioned assertions over retained original executions, with explicit missing evidence | | Model lifecycle | `@bijection/model-lifecycle` | Existing external training/evaluation jobs and governed promotion | | Channel assistants | `@bijection/assistants/channels` | Finite continuation of an authenticated, retained channel occurrence | These source entry points are selected by the platform's immutable component lock. A source import or a unit test does not establish package publication, native deployment compatibility or hosted qualification. ## Start with a metered handle Declare a catalog once with `defineLanguageModels`, install it on the language models component, and pass typed handles to consumers. A handle contains model identity and capabilities; it never contains a reusable provider credential. Credentials remain deployment configuration or native gateway custody. `defineAiTask` combines an input schema, an output schema, instructions and a context builder. It opens a metered session before the builder reads protected records. Each retained intent has one provider attempt, with SDK retries disabled. Results distinguish completed output, invalid input/output, refusal, failure, insufficient evidence, uncertainty and cancellation. An uncertain attempt retains its reservation and must not be replayed to guess whether the provider ran. `defineManagedAgent` adds a declared tool allowlist and bounded steps, tool calls, prompt bytes and output bytes. It checks the native session around tool work. Hitting a step or output limit is a refused outcome, never a completed answer. Tools execute through the application's current authenticated context. ## Publish knowledge from canonical sources `defineKnowledgeCollection` embeds a bounded source population into a namespace isolated by collection and generation. The host owns accepted indexing work and publication. Its publication mutation must recheck the original work, source revisions and previous pointer before accepting ready index entries. Retrieval uses vector results only to select candidates. It rereads canonical sources and their current revisions before returning text and citations, then rechecks the published generation. Cached text cannot bypass a revoked or changed source. Citations identify collection, source, revision, generation and chunk. Different chunks or revisions remain distinct in the workspace. ## Review business changes through native operations The assistant's operation tool creates drafts for a person to submit through the normal operation dialog. Model output does not grant permission or execute a business change. The actual operation rechecks current data and its review rule. Documents use the same boundary: an external worker claims an existing native attempt and reads files through scoped native delivery. Parsed pages and model values retain source evidence. `requireResult` checks the exact definition, current source selection, unresolved issues and required review before a business operation can use the result. The agent workspace exposes host-owned original work, immutable definition identity, model attempts and verified native request links. Cancellation fences late content writes through `/storage`; settlement can still resolve an already dispatched attempt. Recovery reads original retained evidence. The knowledge `/workspace` and `/application` adapters expose source coverage, current publication, bounded refresh history and explicit metered previews. Inspection performs no inference. Current-source checks still apply to previews. ## Measure original executions `defineAiEvaluation` evaluates explicit case/candidate/repetition slots against original outputs, tool traces, citations, latency, cost and optional judgments. Declarations pin target/configuration, case-set revision, baseline, evaluator versions, repetitions, populations and finite cost/deadline limits. Missing or unknown slots remain in the denominator. Duplicate execution IDs cannot count as independent repetitions. Partial cost stays partial, with exact picodollar values retained. `defineAiJudge` from `/ai/judge` runs a pinned rubric through a metered typed task. Its original model-produced judgment has its own uncertainty and accounting. It grants no approval. The evaluation workspace adapters expose versioned definitions and retained reports. `aiLifecycleMeasurements` converts a report into metric/table fragments for the existing model-lifecycle owner, using an explicit selection or held-out validation population. Aggregate cost/deadline status remains explicit. It creates neither a native report identity nor promotion approval. Run new inference as an explicit admitted intent; evaluating stored evidence does not silently run a model. ## Run the reference journeys The repository's [AI operations reference](https://github.com/bijectionhq/bijection/tree/main/examples/ai-operations) connects retained typed classification, support investigation, knowledge refresh, reviewed delivery remedies and refunds, external invoice parsing, document review and evaluation reports in the standard workspace. Its default models are labelled fixtures that call no provider and demonstrate contracts rather than model quality. The optional Slack worker reuses the existing signed channel transport. It accepts the occurrence before inference, resumes only its original request ID, and commits a reply from stored settled evidence. Its separate native service grant admits policy reads; a Slack actor does not acquire a person's finance permissions. Accepted reply submission and confirmed provider delivery remain separate outcomes. The reference also connects original evaluation evidence to native assessment, reviewed promotion, prediction-to-decision creation evidence and observed outcome feedback. Its finite Workflow recipes observe original owners using an explicitly authorized native service credential. They preserve original operation requests and never impersonate the initiating person or repeat inference during replay. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.