Catalogue
Every adapter is open: a trained model you can call through the API, or download and run yourself.
Need a decision that is not here? Request an adapter.
34 adapters
Reasoning and knowledge
Whether evidence supports a claim, two statements conflict or mean the same, an item meets a requirement or a pack is complete; and general-knowledge questions.
Claim support check
jeff.reasoning.entails
Checks whether supplied documents, messages or records support a claim. Returns supported, unsupported or unclear.
Contradiction check
jeff.reasoning.contradicts
Checks whether two statements or records conflict. Returns yes, no or unclear when the information is ambiguous.
Document completeness check
jeff.reasoning.is-complete
Checks a set of documents against a checklist. Returns complete, incomplete or unclear, accounting for missing, expired, unsigned or incorrectly attributed documents.
Multiple-choice general knowledge
jeff.knowledge.knowledge-qa
Selects an answer to a multiple-choice question about science, common sense or general knowledge.
Requirement check
jeff.reasoning.satisfies
Checks whether a document, message or record meets a single stated requirement. Returns pass, fail or unclear.
Same meaning check
jeff.reasoning.equivalent
Checks whether two items have the same meaning within a defined scope, allowing for differences in wording, units and format. Returns yes, no or unclear.
Search and documents
Which source a query goes to, whether a passage or a set of passages answers it, what a document or section is, and whether an extracted value is right.
Document classification
jeff.retrieval.doc-type
Classifies a document using your list of document types and descriptions, or returns other when none fits.
Document section classification
jeff.retrieval.section-id
Identifies the type of document section a passage belongs to, using your list of section types and descriptions. Returns none when no type fits.
Evidence sufficiency check
jeff.retrieval.evidence-sufficient
Checks whether retrieved passages provide enough evidence to answer a question. Returns sufficient, insufficient or conflicting.
Extracted value validation
jeff.retrieval.extract-check
Checks an extracted field and value against the source document. Identifies correct values, incorrect values, wrong-field errors, unsupported values and ambiguity.
Passage relevance
jeff.retrieval.passage-relevant
Checks whether a passage answers a question about the specified subject and time period. Returns yes, partly or no.
Passage selection and answer verification
jeff.retrieval.ground
Selects the passage that best answers a question and checks whether a generated answer is supported by its sources. These are separate functions.
Search source selection
jeff.retrieval.query-route
Selects which document collection, database or tool should handle a query from a menu of sources. Can also indicate that several sources or no source are needed.
AI agents
The checks around each step of an agent: which tool to call, whether an action may run, whether a result helps, what to do next, whether the task is done, and whether a reply needs careful thinking.
Agent action check
jeff.agents.action-check
Checks a proposed tool call against the user's request and the application's rules. Determines whether to approve it, reject it as incorrect or unsafe, or ask the user first.
Agent next-step selection
jeff.agents.after-result
Determines what an agent should do after a tool returns a result: continue, retry, change approach, stop or hand over to a person.
Agent task completion
jeff.agents.task-complete
Checks the user's request against the agent's tool calls, results and final answer. Classifies the task as complete, incomplete or unclear.
Agent tool selection from a menu
jeff.agents.tool-route
Selects the next tool from a menu included in the request, using the conversation for context. Can also indicate that no tool is needed or that the user must supply missing information.
Assistant reasoning check
jeff.agents.assistant-routing
Predicts whether an assistant's next response requires careful reasoning or can be answered directly. Use it to route simpler turns to a faster, cheaper response path.
Tool result relevance
jeff.agents.result-relevant
Checks whether a tool result answers the agent's current question: yes, partly or no.
Routing
Which tool an agent calls next, or which specialist adapter answers a request.
Agent tool selection from labelled options
jeff.routing.tools
Selects the next tool from tools supplied as labelled options with descriptions. Can also indicate that the agent should answer directly or ask the user for missing information.
Open-adapter router
jeff.routing.router-open
Selects an open specialist from your menu to handle a request. Returns base when no specialist fits and the general model should handle it.
Support and inboxes
What an incoming message or ticket asks for, who should handle it, how urgent it is and whether a team must act.
Customer request classification
jeff.support.support-intents
Identifies the request in a customer or assistant-user message, choosing from a fixed list such as cancel_order or track_refund.
Shared inbox triage
jeff.support.message-triage
Classifies an inbox message, determines the required action and urgency, and selects the responsible team using your message categories, routing table and rules.
Support ticket triage
jeff.support.triage
Routes a support ticket or email to a team and assesses urgency, sentiment and the need for human review.
Team action and escalation check
jeff.support.requires-action
Checks a message against a team's policy to determine whether action or escalation is required. Each decision is yes, no or unclear.
Safety
Prompt injection and jailbreak attempts, spam and phishing in what users and outside sources send.
Prompt injection detection
jeff.safety.guard
Detects attempts to override an AI model's instructions, bypass its safeguards or extract data. Also classifies the type of attempt.
Spam and phishing detection
jeff.safety.spam
Detects spam or phishing in text messages and emails. A separate function classifies each message as legitimate, spam or phishing.
Language
The feeling expressed in a short comment or message.
Security operations
What a security team should do about a new alert under its playbook.
Software engineering
What kind of failure a broken build is, and which part of a project owns a new issue or pull request.
Build failure triage
jeff.engineering.ci-failure
Classifies a failed build or test run using its log, code changes and run history. Distinguishes regressions, flaky tests, infrastructure issues, dependency changes and tests that need updating.
Issue and pull request classification
jeff.engineering.issue-route
Assigns an issue or pull request to a project component and classifies it as a bug, feature, question, documentation change or duplicate.
Compliance
Applying a firm's rules: sanctions screening of names.
Legal
What type a contract clause is.
Trading
What a trading desk's rules say to do with the order book.