feat(memory): allow disabling built-in memory toolfeat(memory): allow disabling built-in memory tool

[MyQiongbao]

Closes #39492

What does this PR do?

Adds memory.builtin_tool.enabled to allow deployments to hide the built-in global memory tool while keeping external memory providers enabled.

memory:
  builtin_tool:
    enabled: false

Default behavior remains unchanged: the built-in memory tool is still exposed unless this option is explicitly set to false.

This is useful for multi-user API/gateway deployments where external memory providers can scope memory per session/user, but the built-in tool writes to
global MEMORY.md and USER.md.

## Related Issue

Closes #39492

## Type of Change

- [ ] 🐛 Bug fix (non-breaking change that fixes an issue)
- [x] ✨ New feature (non-breaking change that adds functionality)
- [ ] 🔒 Security fix
- [x] 📝 Documentation update
- [x] ✅ Tests (adding or improving test coverage)
- [ ] ♻️ Refactor (no behavior change)
- [ ] 🎯 New skill (bundled or hub)

## Changes Made

- agent/memory_config.py: add helper for reading memory.builtin_tool.enabled
- model_tools.py: filter the built-in memory tool from tool definitions when disabled
- agent/agent_init.py: record the built-in memory tool config on the agent
- hermes_cli/config.py: add the default config value
- cli-config.yaml.example: document the new config key
- website/docs/user-guide/configuration.md: document the new config key
- website/docs/user-guide/features/memory.md: document the new config key
- tests/test_model_tools.py: add coverage for default enabled and explicit disabled behavior
- tests/run_agent/test_memory_provider_init.py: verify disabling the built-in tool does not disable external provider tools

## How to Test

1. Run targeted model/tool initialization tests:

uv run --extra dev python -m pytest tests/test_model_tools.py tests/run_agent/test_memory_provider_init.py tests/tools/test_memory_tool_schema.py -q -o
addopts=

2. Run memory provider tests:

uv run --extra dev python -m pytest tests/agent/test_memory_provider.py -q -o addopts=

3. Run lint/checks:

uv run --extra dev ruff check agent/memory_config.py model_tools.py agent/agent_init.py tests/test_model_tools.py tests/run_agent/
test_memory_provider_init.py
git diff --check

## Checklist

### Code

- [x] I've read the Contributing Guide (https://github.com/NousResearch/hermes-agent/blob/main/CONTRIBUTING.md)
- [x] My commit messages follow Conventional Commits (https://www.conventionalcommits.org/) (fix(scope):, feat(scope):, etc.)
- [x] I searched for existing PRs (https://github.com/NousResearch/hermes-agent/pulls) to make sure this isn't a duplicate
- [x] My PR contains only changes related to this fix/feature (no unrelated commits)
- [ ] I've run pytest tests/ -q and all tests pass
- [x] I've added tests for my changes (required for bug fixes, strongly encouraged for features)
- [x] I've tested on my platform: Linux container

### Documentation & Housekeeping

- [x] I've updated relevant documentation (README, docs/, docstrings) — or N/A
- [x] I've updated cli-config.yaml.example if I added/changed config keys — or N/A
- [x] I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
- [x] I've considered cross-platform impact (Windows, macOS) per the compatibility guide
  (https://github.com/NousResearch/hermes-agent/blob/main/CONTRIBUTING.md#cross-platform-compatibility) — or N/A

- [x] I've updated tool descriptions/schemas if I changed tool behavior — or N/A

## Screenshots / Logs

N/A

[teknium1]

Thanks for isolating the built-in schema from external provider tools. The request remains relevant on current main, but the current patch needs a scope decision before salvage.

Problems

• The schema filter does not stop the global built-in store from reaching the model. Current main loads it when memory_enabled or user_profile_enabled is true (agent/agent_init.py:1363-1375) and injects its MEMORY.md / USER.md snapshots independently of tool availability (agent/system_prompt.py:460-469). That leaves the multi-user isolation rationale incomplete.
• agent._builtin_memory_tool_enabled is assigned in the proposed agent/agent_init.py:1065, but the proposed filter independently reloads config in model_tools.py; the stored value has no reader.
• The new provider test omits the built-in-memory enable flags and does not test prompt injection, so it misses the deployment state described in the PR.

Suggested changes

• Decide whether this flag is schema-only or disables the built-in memory subsystem; if it is for isolation, gate the global-store prompt path too while preserving provider tools.
• Add a temp-HERMES_HOME integration test covering provider tools, built-in schema visibility, and built-in prompt blocks.

Automated hermes-sweeper review.

[teknium1]

This removes only the API schema. With memory_enabled or user_profile_enabled still true, current main loads the global store in agent/agent_init.py:1363-1375 and injects its snapshots in agent/system_prompt.py:460-469; that does not fully address the multi-user isolation rationale. Please define whether this flag is schema-only or must also gate built-in prompt/store behavior.

[teknium1]

This attribute is not read by this PR: the schema path independently reloads config in model_tools.py. Please remove it or use one shared decision path so the new configuration cannot silently diverge between agent state and tool construction.

[teknium1]

Fixed on main via #90559 (salvage of #90413). You were among the earliest (Jun 5) to propose hiding the built-in memory tool while keeping external providers — thank you, credited here. The landed approach gates via the existing check_fn on the store flags (memory_enabled / user_profile_enabled) rather than a new builtin_tool.enabled key, so no new config surface was needed. Closing as covered on main.