SurfSense/surfsense_backend/app/automations/actions/agent_task/params.py

53 lines
1.9 KiB
Python
Raw Normal View History

"""``AgentTaskActionParams`` — params for the ``agent_task`` action type."""
feat(automation): add Pydantic schemas for the automation definition Three layers of Pydantic models under app/automations/schemas/, one file per concern (SRP), matching the envelope in automation-design-plan.md §5. definition/ — the editable envelope persisted in automations.definition: - envelope.py AutomationDefinition (top-level shape) - plan_step.py PlanStep (one step in the sequential plan) - inputs.py InputsBlock (the inputs JSON Schema wrapper) - execution.py ExecutionBlock (timeouts, retries, concurrency, budget cap, on_failure plan) - metadata.py MetadataBlock (tags + created_from_nl + extras) - trigger_spec.py TriggerSpec (one entry in triggers[]) triggers/ — per-trigger config schemas, dispatched by registry on the TriggerSpec.type discriminator: - schedule.py ScheduleTriggerConfig(cron, timezone) - manual.py ManualTriggerConfig() — empty in v1 actions/ — per-action config schemas, dispatched by registry on the PlanStep.action discriminator: - agent_task.py AgentTaskActionConfig(prompt, tools, model, output_schema) Design properties verified by an inline smoke test: - The §5 worked example round-trips through model_validate_json / model_dump_json byte-for-byte (InputsBlock uses serialize_by_alias so the JSON key stays "schema" not "schema_"). - Envelope rejects unknown top-level keys (extra="forbid"). - MetadataBlock tolerates unknown keys (extra="allow"). - ExecutionBlock defaults apply when the block is omitted. - retry_backoff and concurrency are typed as Literal — bogus values rejected at validation time. - Per-type configs enforce their required fields (cron + timezone on schedule; non-empty prompt on agent_task). The envelope keeps trigger and action configs as untyped dicts on purpose — per-type validation is a registry-driven dispatch (commit 10), keeping the envelope free of every-type-knows-every-type coupling.
2026-05-26 22:50:52 +02:00
from __future__ import annotations
from pydantic import BaseModel, ConfigDict, Field
from app.schemas.new_chat import MentionedDocumentInfo
feat(automation): add Pydantic schemas for the automation definition Three layers of Pydantic models under app/automations/schemas/, one file per concern (SRP), matching the envelope in automation-design-plan.md §5. definition/ — the editable envelope persisted in automations.definition: - envelope.py AutomationDefinition (top-level shape) - plan_step.py PlanStep (one step in the sequential plan) - inputs.py InputsBlock (the inputs JSON Schema wrapper) - execution.py ExecutionBlock (timeouts, retries, concurrency, budget cap, on_failure plan) - metadata.py MetadataBlock (tags + created_from_nl + extras) - trigger_spec.py TriggerSpec (one entry in triggers[]) triggers/ — per-trigger config schemas, dispatched by registry on the TriggerSpec.type discriminator: - schedule.py ScheduleTriggerConfig(cron, timezone) - manual.py ManualTriggerConfig() — empty in v1 actions/ — per-action config schemas, dispatched by registry on the PlanStep.action discriminator: - agent_task.py AgentTaskActionConfig(prompt, tools, model, output_schema) Design properties verified by an inline smoke test: - The §5 worked example round-trips through model_validate_json / model_dump_json byte-for-byte (InputsBlock uses serialize_by_alias so the JSON key stays "schema" not "schema_"). - Envelope rejects unknown top-level keys (extra="forbid"). - MetadataBlock tolerates unknown keys (extra="allow"). - ExecutionBlock defaults apply when the block is omitted. - retry_backoff and concurrency are typed as Literal — bogus values rejected at validation time. - Per-type configs enforce their required fields (cron + timezone on schedule; non-empty prompt on agent_task). The envelope keeps trigger and action configs as untyped dicts on purpose — per-type validation is a registry-driven dispatch (commit 10), keeping the envelope free of every-type-knows-every-type coupling.
2026-05-26 22:50:52 +02:00
class AgentTaskActionParams(BaseModel):
"""Run a multi_agent_chat turn from an automation step."""
feat(automation): add Pydantic schemas for the automation definition Three layers of Pydantic models under app/automations/schemas/, one file per concern (SRP), matching the envelope in automation-design-plan.md §5. definition/ — the editable envelope persisted in automations.definition: - envelope.py AutomationDefinition (top-level shape) - plan_step.py PlanStep (one step in the sequential plan) - inputs.py InputsBlock (the inputs JSON Schema wrapper) - execution.py ExecutionBlock (timeouts, retries, concurrency, budget cap, on_failure plan) - metadata.py MetadataBlock (tags + created_from_nl + extras) - trigger_spec.py TriggerSpec (one entry in triggers[]) triggers/ — per-trigger config schemas, dispatched by registry on the TriggerSpec.type discriminator: - schedule.py ScheduleTriggerConfig(cron, timezone) - manual.py ManualTriggerConfig() — empty in v1 actions/ — per-action config schemas, dispatched by registry on the PlanStep.action discriminator: - agent_task.py AgentTaskActionConfig(prompt, tools, model, output_schema) Design properties verified by an inline smoke test: - The §5 worked example round-trips through model_validate_json / model_dump_json byte-for-byte (InputsBlock uses serialize_by_alias so the JSON key stays "schema" not "schema_"). - Envelope rejects unknown top-level keys (extra="forbid"). - MetadataBlock tolerates unknown keys (extra="allow"). - ExecutionBlock defaults apply when the block is omitted. - retry_backoff and concurrency are typed as Literal — bogus values rejected at validation time. - Per-type configs enforce their required fields (cron + timezone on schedule; non-empty prompt on agent_task). The envelope keeps trigger and action configs as untyped dicts on purpose — per-type validation is a registry-driven dispatch (commit 10), keeping the envelope free of every-type-knows-every-type coupling.
2026-05-26 22:50:52 +02:00
model_config = ConfigDict(extra="forbid")
query: str = Field(
...,
min_length=1,
description="User query for the agent; rendered at execute time.",
feat(automation): add Pydantic schemas for the automation definition Three layers of Pydantic models under app/automations/schemas/, one file per concern (SRP), matching the envelope in automation-design-plan.md §5. definition/ — the editable envelope persisted in automations.definition: - envelope.py AutomationDefinition (top-level shape) - plan_step.py PlanStep (one step in the sequential plan) - inputs.py InputsBlock (the inputs JSON Schema wrapper) - execution.py ExecutionBlock (timeouts, retries, concurrency, budget cap, on_failure plan) - metadata.py MetadataBlock (tags + created_from_nl + extras) - trigger_spec.py TriggerSpec (one entry in triggers[]) triggers/ — per-trigger config schemas, dispatched by registry on the TriggerSpec.type discriminator: - schedule.py ScheduleTriggerConfig(cron, timezone) - manual.py ManualTriggerConfig() — empty in v1 actions/ — per-action config schemas, dispatched by registry on the PlanStep.action discriminator: - agent_task.py AgentTaskActionConfig(prompt, tools, model, output_schema) Design properties verified by an inline smoke test: - The §5 worked example round-trips through model_validate_json / model_dump_json byte-for-byte (InputsBlock uses serialize_by_alias so the JSON key stays "schema" not "schema_"). - Envelope rejects unknown top-level keys (extra="forbid"). - MetadataBlock tolerates unknown keys (extra="allow"). - ExecutionBlock defaults apply when the block is omitted. - retry_backoff and concurrency are typed as Literal — bogus values rejected at validation time. - Per-type configs enforce their required fields (cron + timezone on schedule; non-empty prompt on agent_task). The envelope keeps trigger and action configs as untyped dicts on purpose — per-type validation is a registry-driven dispatch (commit 10), keeping the envelope free of every-type-knows-every-type coupling.
2026-05-26 22:50:52 +02:00
)
auto_approve_all: bool = Field(
default=False,
description="If true, every HITL approval is auto-approved; otherwise rejected.",
feat(automation): add Pydantic schemas for the automation definition Three layers of Pydantic models under app/automations/schemas/, one file per concern (SRP), matching the envelope in automation-design-plan.md §5. definition/ — the editable envelope persisted in automations.definition: - envelope.py AutomationDefinition (top-level shape) - plan_step.py PlanStep (one step in the sequential plan) - inputs.py InputsBlock (the inputs JSON Schema wrapper) - execution.py ExecutionBlock (timeouts, retries, concurrency, budget cap, on_failure plan) - metadata.py MetadataBlock (tags + created_from_nl + extras) - trigger_spec.py TriggerSpec (one entry in triggers[]) triggers/ — per-trigger config schemas, dispatched by registry on the TriggerSpec.type discriminator: - schedule.py ScheduleTriggerConfig(cron, timezone) - manual.py ManualTriggerConfig() — empty in v1 actions/ — per-action config schemas, dispatched by registry on the PlanStep.action discriminator: - agent_task.py AgentTaskActionConfig(prompt, tools, model, output_schema) Design properties verified by an inline smoke test: - The §5 worked example round-trips through model_validate_json / model_dump_json byte-for-byte (InputsBlock uses serialize_by_alias so the JSON key stays "schema" not "schema_"). - Envelope rejects unknown top-level keys (extra="forbid"). - MetadataBlock tolerates unknown keys (extra="allow"). - ExecutionBlock defaults apply when the block is omitted. - retry_backoff and concurrency are typed as Literal — bogus values rejected at validation time. - Per-type configs enforce their required fields (cron + timezone on schedule; non-empty prompt on agent_task). The envelope keeps trigger and action configs as untyped dicts on purpose — per-type validation is a registry-driven dispatch (commit 10), keeping the envelope free of every-type-knows-every-type coupling.
2026-05-26 22:50:52 +02:00
)
# @-mention references chosen in the task input. Mirror the ``new_chat``
# request fields (minus SurfSense product docs) so the run can scope
# retrieval to the user's selected files / folders / connectors. All
# optional and additive; a task with no mentions behaves as before.
mentioned_document_ids: list[int] | None = Field(
default=None,
description="Knowledge-base document IDs the task references with @.",
)
mentioned_folder_ids: list[int] | None = Field(
default=None,
description="Knowledge-base folder IDs the task references with @.",
)
mentioned_connector_ids: list[int] | None = Field(
default=None,
description="Concrete connector account IDs the task references with @.",
)
mentioned_connectors: list[MentionedDocumentInfo] | None = Field(
default=None,
description="Display/context metadata for the @-mentioned connector accounts.",
)
mentioned_documents: list[MentionedDocumentInfo] | None = Field(
default=None,
description=(
"Chip metadata (id, title, kind, ...) for every @-mention so the "
"run can resolve titles to virtual paths and substitute them in "
"the query."
),
)