refactor: move agent template rendering from model to core layer #929

hyacinthus · 2025-12-29T09:44:24Z

Changes

Created new function in that handles template rendering when exists
Deprecated method with deprecation warning for future removal
Replaced all calls outside of models/ directory with new function
Maintained backward compatibility by keeping for internal model usage

Migration Path

Template rendering logic moved from model layer to core layer:

Before: Template rendering was done at model level
After: retrieves agent from DB and applies template if is set

All consuming code now uses from core layer, maintaining the same functionality with cleaner separation of concerns.

Related to template feature development.

Copilot

Pull request overview

This PR refactors agent template rendering by moving the logic from the model layer to a new core layer. The main goal is to achieve cleaner separation of concerns by having template rendering handled at the application/core level rather than within the ORM models.

Key Changes:

New get_agent() function in intentkit/core/agent.py that retrieves agents and applies template rendering
New intentkit/core/template.py module with render_agent() and create_template_from_agent() functions
Deprecated Agent.get() method with a deprecation warning pointing to the new function
Replaced all external calls to Agent.get() with the new get_agent() function across 10+ files
Added new database columns template_id and extra_prompt to AgentTable
Created new Template and TemplateTable models extending AgentCore

Reviewed changes

Copilot reviewed 16 out of 16 changed files in this pull request and generated 8 comments.

Show a summary per file

File	Description
`intentkit/models/template.py`	New template models (`Template`, `TemplateTable`) that extend `AgentCore` with template-specific fields
`intentkit/models/agent.py`	Added `template_id` and `extra_prompt` columns to `AgentTable`; added `from_template_id` and `extra_prompt` fields to `AgentCreate`; deprecated `Agent.get()` method; moved `short_term_memory_strategy` into `AgentCore`
`intentkit/core/template.py`	New module with template rendering logic: `render_agent()` applies template fields to agents, `create_template_from_agent()` creates templates from agents
`intentkit/core/agent.py`	New `get_agent()` function that retrieves agents and applies template rendering if `template_id` is set
`intentkit/core/manager/service.py`	Migrated from `Agent.get()` to `get_agent()`
`intentkit/core/engine.py`	Migrated from `Agent.get()` to `get_agent()` in multiple locations
`intentkit/core/draft.py`	Migrated from `Agent.get()` to `get_agent()`
`intentkit/core/credit.py`	Migrated from `Agent.get()` to `get_agent()`
`intentkit/core/asset.py`	Migrated from `Agent.get()` to `get_agent()`
`app/services/twitter/oauth2_callback.py`	Migrated from `Agent.get()` to `get_agent()`
`app/entrypoints/web.py`	Migrated from `Agent.get()` to `get_agent()` in 9 endpoint handlers
`app/entrypoints/openai_compatible.py`	Migrated from `Agent.get()` to `get_agent()`
`app/entrypoints/agent_api.py`	Migrated from `Agent.get()` to `get_agent()` in 7 endpoint handlers
`app/admin/api.py`	Migrated from `Agent.get()` to `get_agent()` (aliased as `get_agent_by_id`) in 6 locations
`intentkit/utils/slack_alert.py`	Added explicit `-> None` return type annotation; changed to discard response instead of returning it
`CHANGELOG.md`	Added v0.8.37 changelog entry (appears unrelated to this PR)

Comments suppressed due to low confidence (10)

app/admin/api.py:80

This comment appears to contain commented-out code.

    # if not input.owner:
    #     raise IntentKitAPIError(
    #         status_code=400, key="BadRequest", message="Owner is required"

app/admin/api.py:88

This comment appears to contain commented-out code.

    # if user_id:
    #     if input.owner != user_id:
    #         raise IntentKitAPIError(
    #             status_code=400,
    #             key="BadRequest",
    #             message="Owner does not match user ID",

app/admin/api.py:95

This comment appears to contain commented-out code.

    # if user:
    #     max_fee += user.nft_count * 10
    # if input.fee_percentage and input.fee_percentage > max_fee:
    #     raise IntentKitAPIError(
    #         status_code=400, key="BadRequest", message="Fee percentage too high"

app/admin/api.py:130

This comment appears to contain commented-out code.

    # if not input.owner:
    #     raise IntentKitAPIError(
    #         status_code=400, key="BadRequest", message="Owner is required"

app/admin/api.py:138

This comment appears to contain commented-out code.

    # if user_id:
    #     if input.owner != user_id:
    #         raise IntentKitAPIError(
    #             status_code=400,
    #             key="BadRequest",
    #             message="Owner does not match user ID",

app/admin/api.py:657

Variable latest_agent is not used.

    latest_agent, agent_data = await deploy_agent(agent_id, agent, "admin")

app/admin/api.py:657

Variable agent_data is not used.

    latest_agent, agent_data = await deploy_agent(agent_id, agent, "admin")

intentkit/core/engine.py:560

Variable last is not used.

                    last = this_time

intentkit/core/engine.py:808

Variable last is not used.

                        last = this_time

intentkit/models/agent.py:752

This import of module re is redundant, as it was previously imported on line 6.

        import re

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2025-12-29T09:50:47Z

intentkit/utils/slack_alert.py

    thread_ts: str | None = None,
    channel: str | None = None,
-):
+) -> None:


The function signature change from returning an implicit None to explicitly declaring -> None is good, but this also removes the ability for callers to check if the message was successfully sent by examining the response. The previous code returned the Slack API response which could be useful for debugging or confirmation. If callers don't need this information, this change is fine, but it's a breaking change for any code that uses the return value.

Copilot · 2025-12-29T09:50:48Z

intentkit/models/agent.py


-    short_term_memory_strategy: Annotated[
-        Literal["trim", "summarize"] | None,
+    # only when wallet privder is readonly


Spelling error in comment: "privder" should be "provider".

Suggested change

# only when wallet privder is readonly

# only when wallet provider is readonly

Copilot · 2025-12-29T09:50:48Z

intentkit/models/agent.py

+    short_term_memory_strategy: Annotated[
+        Literal["trim", "summarize"] | None,
        PydanticField(
-            default=None,
-            description="Provider of the agent's wallet",
+            default="trim",
+            description="Strategy for managing short-term memory when context limit is reached. 'trim' removes oldest messages, 'summarize' creates summaries.",
        ),
    ]


The short_term_memory_strategy field was moved from AgentUserInput to AgentCore. This means it will now be included in template rendering and agent hash calculations. This is a significant behavioral change that should be explicitly mentioned in the PR description and tested. Verify that this is the intended behavior, as it changes the scope of which fields are considered "core" to an agent's identity.

Copilot · 2025-12-29T09:50:48Z

intentkit/models/agent.py

+        String,
+        nullable=True,
+        comment="Only when the agent is created from a template.",
+        max_length=20000,


The max_length=20000 parameter is not valid for SQLAlchemy's mapped_column. SQLAlchemy String columns should use String(20000) to specify the length limit. The current code will create a column without a length constraint, which may cause issues when trying to insert strings longer than the database default.

Suggested change

String,

nullable=True,

comment="Only when the agent is created from a template.",

max_length=20000,

String(20000),

nullable=True,

comment="Only when the agent is created from a template.",

Copilot · 2025-12-29T09:50:48Z

intentkit/models/agent.py

+    from_template_id: Annotated[
+        str | None,
+        PydanticField(
+            default=None,
+            description="Template identifier of the agent",
+            max_length=50,
+        ),
+    ]


The field name mismatch between from_template_id in the Pydantic model and template_id in the SQLAlchemy table will cause the template_id to not be saved when creating agents. When model_dump() is called in the create() method, it will produce a dictionary with key from_template_id, but the AgentTable expects template_id. Consider either:

Renaming the Pydantic field to template_id to match the database column

Using Pydantic's Field(alias="template_id") with by_alias=True in model_dump

Manually mapping the field before passing to AgentTable

Copilot · 2025-12-29T09:50:49Z

intentkit/core/template.py

+    # Get template_id from the agent
+    # Since Agent model may not have template_id mapped, we query from DB
+    async with get_session() as db:
+        result = await db.execute(
+            select(AgentTable.template_id).where(AgentTable.id == agent.id)
+        )
+        row = result.first()
+        if row is None:
+            return agent
+        template_id = row[0]


The render_agent function queries the database twice for the same agent - once to get the template_id and again in get_agent. This is inefficient. Consider passing the template_id as a parameter to render_agent or accepting the full AgentTable item to avoid the redundant database query at lines 71-77.

Copilot · 2025-12-29T09:50:49Z

CHANGELOG.md

+## v0.8.37 - 2025-01-27
+
+### Features
+- Frontend skill box display in chat interface
+- Local frontend development improvements
+- GPT image model updated to version 1.5
+
+### Fixes
+- Docker compose configuration fixes
+- Debug authentication removed
+- Various lint fixes
+
+### Chores
+- Dependency updates
+
+**Full Changelog**: https://github.com/crestalnetwork/intentkit/compare/v0.8.36...v0.8.37


The CHANGELOG entry appears to be unrelated to this PR. This PR is about refactoring agent template rendering, but the CHANGELOG describes frontend features, Docker fixes, and other unrelated changes. Either the CHANGELOG entry should be removed from this PR or it needs to be updated to reflect the actual changes made in this refactoring.

hyacinthus added 7 commits December 29, 2025 01:09

doc: changelog

79ca24e

feat: extra prompt with template

80edb22

feat: template model

5a42e2e

feat: template logic

79f031b

refactor: remove return value from send_slack_message function

fcc7834

fix: lint in slack

2090be7

refactor: move agent template rendering from model to core layer

0a32528

Copilot AI review requested due to automatic review settings December 29, 2025 09:44

hyacinthus merged commit b6a8632 into main Dec 29, 2025
5 checks passed

hyacinthus deleted the hyacinthus branch December 29, 2025 09:44

Copilot started reviewing on behalf of hyacinthus December 29, 2025 09:44 View session

Copilot AI reviewed Dec 29, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

refactor: move agent template rendering from model to core layer #929

refactor: move agent template rendering from model to core layer #929

Uh oh!

hyacinthus commented Dec 29, 2025

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Dec 29, 2025

Uh oh!

Copilot AI Dec 29, 2025

Uh oh!

Copilot AI Dec 29, 2025

Uh oh!

Copilot AI Dec 29, 2025

Uh oh!

Copilot AI Dec 29, 2025

Uh oh!

Copilot AI Dec 29, 2025

Uh oh!

Copilot AI Dec 29, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	# only when wallet privder is readonly
	# only when wallet provider is readonly

refactor: move agent template rendering from model to core layer #929

refactor: move agent template rendering from model to core layer #929

Uh oh!

Conversation

hyacinthus commented Dec 29, 2025

Changes

Migration Path

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Dec 29, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 29, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 29, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 29, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 29, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 29, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 29, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants