Job Administration¶

job_administration — G6 Administration job agent.

Cluster: Job Agents | Type: component | MCP Tools: 26

Overview¶

Domain-specialist job agent for office and administrative operations. Automates document processing, meeting scheduling, records management, correspondence drafting, regulatory filing, workflow analysis, compliance evaluation, and efficiency assessments — covering the full lifecycle of administrative support within a safety-bounded, audit-trailed JobAgentBlock framework.

When to use:

• Processing and routing incoming documents or correspondence at scale
• Scheduling and coordinating meetings across teams with conflict detection
• Auditing administrative workflows for compliance gaps or inefficiencies
• Generating templated reports and managing structured records in a central store

Example:

from mvp.job_administration import JobAdministrationBlock, JobAdministrationInput

block = JobAdministrationBlock()
result = block.infer(JobAdministrationInput(
    task="Draft a formal response to the vendor's contract renewal notice and file it in records",
    context={"vendor": "Acme Corp", "contract_id": "VC-2024-081"},
))
# result.ok → True; result.value → JobAdministrationOutput with result, artifacts

Works well with: job_framework, job_manager, job_lawyer

Launch Caveat¶

This component is pilot-ready for administrative workflow assistance, not an enterprise records or compliance system. It is suitable for first-customer workflows where a user needs document triage, scheduling support, artifact storage, workflow analysis, and compliance checklists that return quickly and degrade predictably.

Do not treat outputs as legal, regulatory, records-retention, HR, procurement, or safety determinations without qualified human review. Live search and model-backed debate are disabled by default to protect onboarding reliability; enable them explicitly with G6_JOB_ADMINISTRATION_LIVE_INTEGRATIONS=1 when richer analysis is worth the latency and dependency risk. Optional integrations are bounded by G6_JOB_ADMINISTRATION_INTEGRATION_TIMEOUT_SEC or per-call parameters.integration_timeout_sec, and timeout details are surfaced in metadata.integration_warnings.

Architecture note: AdministrationStore is the primary live backend today and persists the component's SQLite-backed documents, schedules, correspondence, procurement, and facilities records. ERPAdapter is an optional passthrough/scaffold abstraction for future ERP integration; the default InMemoryERPAdapter is simulation-only and must not be mistaken for SAP, Oracle, Dynamics, or any other production ERP connector. The current MCP surface reports this status in get_capabilities and does not route write-side ERP methods without explicit review.

Public API¶

JobAdministrationBlock(JobAgentBlock)¶

G6 Administration job agent — Tier 1 block.

Methods:

infer(data: object) -> Result[JobAdministrationOutput]¶

JobAdministrationInput(JobInput)¶

Input for the Administration job agent.

JobAdministrationOutput(JobOutput)¶

JobAdministrationMCPBlock(AIBlock[MCPJobAdministrationInput, MCPJobAdministrationOutput, dict])¶

26-op MCP block for the Administration job agent.

Methods:

infer(data: MCPJobAdministrationInput) -> Result[MCPJobAdministrationOutput]¶

MCPJobAdministrationInput(BaseModel)¶

Input to JobAdministrationMCPBlock — 26-op dispatch.

MCPJobAdministrationOutput(BaseModel)¶

Output from JobAdministrationMCPBlock.

AdministrationStore(JobStore)¶

SQLite store for the Administration job agent.

Constructor:

Methods:

create_document(title: str, doc_type: str = '', category: str = '', author: str = '', content: dict | None = None, status: str = 'draft') -> str¶

get_document(doc_id: str) -> dict | None¶

list_documents(category: str = '', status: str = '') -> list[dict]¶

update_document_status(doc_id: str, status: str) -> bool¶

increment_document_version(doc_id: str) -> int¶

create_schedule(title: str, event_type: str = '', organizer: str = '', attendees: list[str] | None = None, start_time: str = '', end_time: str = '', location: str = '', recurrence: dict | None = None) -> str¶

get_schedule(schedule_id: str) -> dict | None¶

list_schedules(event_type: str = '', status: str = '') -> list[dict]¶

update_schedule_status(schedule_id: str, status: str) -> bool¶

create_correspondence(direction: str = 'outbound', counterparty: str = '', subject: str = '', body: str = '', attachments: list[str] | None = None, priority: str = 'normal') -> str¶

get_correspondence(corr_id: str) -> dict | None¶

list_correspondence(direction: str = '', priority: str = '', status: str = '') -> list[dict]¶

update_correspondence_status(corr_id: str, status: str) -> bool¶

create_order(vendor: str, items: list[dict] | None = None, total_amount: float = 0, data: dict | None = None) -> str¶

get_order(order_id: str) -> dict | None¶

list_orders(vendor: str = '', approval_status: str = '') -> list[dict]¶

update_order_approval(order_id: str, approval_status: str) -> bool¶

mark_order_received(order_id: str) -> bool¶

create_facility(name: str, location: str = '', capacity: int = 0, equipment: list[str] | None = None, maintenance: list[dict] | None = None) -> str¶

get_facility(facility_id: str) -> dict | None¶

list_facilities(status: str = '') -> list[dict]¶

update_facility_status(facility_id: str, status: str) -> bool¶

update_facility_maintenance(facility_id: str, maintenance: list[dict]) -> bool¶

Functions¶

assemble_review_text(output: Any) -> str¶

Collect the reviewable free text from a JobAdministrationOutput (duck-typed).

assess_administration_output(output: Any, qa_block: Any | None = None, generate: Any | None = None) -> GroundedRunResult¶

Run grounded four-valued QA over an administration output.

MCP Tools¶