Timmy_Foundation/hermes-agent
15
0
Fork 0
Code Issues 61 Pull Requests 1 Actions 1 Packages Projects Releases Wiki Activity
Files
145f719d309b1ecdde8417c4c6d2d20184817b21
hermes-agent/TODO.md

130 lines
5.4 KiB
Markdown
Raw Normal View History

Add a claude code-like CLI - Introduced `cli-config.yaml.example` to provide a template for configuring the CLI behavior, including model settings, terminal tool configurations, agent behavior, and toolsets. - Created `cli.py` for an interactive terminal interface, allowing users to start the Hermes Agent with various options and toolsets. - Added `hermes` launcher script for convenient CLI access. - Updated `model_tools.py` to support quiet mode for suppressing output during tool initialization and execution. - Enhanced logging in various tools to respect quiet mode, improving user experience by reducing unnecessary output. - Added `prompt_toolkit` to `requirements.txt` for improved CLI interaction capabilities. - Created `TODO.md` for future improvements and enhancements to the Hermes Agent framework.
2026-01-31 06:30:48 +00:00
# Hermes Agent - Future Improvements
---
Add todo tool for agent task planning and management Single `todo` tool that reads (no params) or writes (provide todos array with merge flag). In-memory TodoStore on AIAgent, no system prompt mutation, behavioral guidance in tool description only. State re-injected after context compression events. Gateway sessions hydrate from conversation history. Added to all platform toolsets. Also wired into RL agent_loop.py with per-run TodoStore and fixed browser_snapshot user_task passthrough from first user message.
2026-02-17 17:02:33 -08:00
feat: update TODO.md with enhancements to skills and memory systems - Increased the tool count to 44+ and clarified the management of bundled and agent-managed skills. - Introduced a persistent memory system with MEMORY.md and USER.md for agent notes and user profiles. - Updated the storage evolution section to reflect the use of SQLite for sessions and clarified the organization of skills and memories. - Added current status of memory types implemented, highlighting progress in agent intelligence capabilities.
2026-02-19 18:47:45 -08:00
## 3. Local Browser Control via CDP 🌐
Update agent configuration for maximum tool-calling iterations - Increased the default maximum tool-calling iterations from 20 to 60 in the CLI configuration and related files, allowing for more complex tasks. - Updated documentation and comments to reflect the new recommended range for iterations, enhancing user guidance. - Implemented backward compatibility for loading max iterations from the root-level configuration, ensuring a smooth transition for existing users. - Adjusted the setup wizard to prompt for the maximum iterations setting, improving user experience during configuration.
2026-02-03 14:48:19 -08:00
Add todo tool for agent task planning and management Single `todo` tool that reads (no params) or writes (provide todos array with merge flag). In-memory TodoStore on AIAgent, no system prompt mutation, behavioral guidance in tool description only. State re-injected after context compression events. Gateway sessions hydrate from conversation history. Added to all platform toolsets. Also wired into RL agent_loop.py with per-run TodoStore and fixed browser_snapshot user_task passthrough from first user message.
2026-02-17 17:02:33 -08:00
**Status:** Not started (currently Browserbase cloud only)
**Priority:** Medium
Support local Chrome/Chromium via Chrome DevTools Protocol alongside existing Browserbase cloud backend.
**What other agents do:**
- **OpenClaw**: Full CDP-based Chrome control with snapshots, actions, uploads, profiles, file chooser, PDF save, console messages, tab management. Uses local Chrome for persistent login sessions.
- **Cline**: Headless browser with Computer Use (click, type, scroll, screenshot, console logs)
**Our approach:**
- Add a `local` backend option to `browser_tool.py` using Playwright or raw CDP
- Config toggle: `browser.backend: local | browserbase | auto`
- `auto` mode: try local first, fall back to Browserbase
- Local advantages: free, persistent login sessions, no API key needed
- Local disadvantages: no CAPTCHA solving, no stealth mode, requires Chrome installed
- Reuse the same 10-tool interface -- just swap the backend
- Later: Chrome profile management for persistent sessions across restarts
---
Update agent configuration for maximum tool-calling iterations - Increased the default maximum tool-calling iterations from 20 to 60 in the CLI configuration and related files, allowing for more complex tasks. - Updated documentation and comments to reflect the new recommended range for iterations, enhancing user guidance. - Implemented backward compatibility for loading max iterations from the root-level configuration, ensuring a smooth transition for existing users. - Adjusted the setup wizard to prompt for the maximum iterations setting, improving user experience during configuration.
2026-02-03 14:48:19 -08:00
feat: update TODO.md with enhancements to skills and memory systems - Increased the tool count to 44+ and clarified the management of bundled and agent-managed skills. - Introduced a persistent memory system with MEMORY.md and USER.md for agent notes and user profiles. - Updated the storage evolution section to reflect the use of SQLite for sessions and clarified the organization of skills and memories. - Added current status of memory types implemented, highlighting progress in agent intelligence capabilities.
2026-02-19 18:47:45 -08:00
## 4. Signal Integration 📡
Update agent configuration for maximum tool-calling iterations - Increased the default maximum tool-calling iterations from 20 to 60 in the CLI configuration and related files, allowing for more complex tasks. - Updated documentation and comments to reflect the new recommended range for iterations, enhancing user guidance. - Implemented backward compatibility for loading max iterations from the root-level configuration, ensuring a smooth transition for existing users. - Adjusted the setup wizard to prompt for the maximum iterations setting, improving user experience during configuration.
2026-02-03 14:48:19 -08:00
Add todo tool for agent task planning and management Single `todo` tool that reads (no params) or writes (provide todos array with merge flag). In-memory TodoStore on AIAgent, no system prompt mutation, behavioral guidance in tool description only. State re-injected after context compression events. Gateway sessions hydrate from conversation history. Added to all platform toolsets. Also wired into RL agent_loop.py with per-run TodoStore and fixed browser_snapshot user_task passthrough from first user message.
2026-02-17 17:02:33 -08:00
**Status:** Not started
**Priority:** Low
Refactor TODO.md to summarize future improvements for the Hermes Agent, focusing on subagent architecture, task management, dynamic skills expansion, and interactive clarifying questions. Key ideas include context isolation for subagents, task decomposition, progress tracking, and skill acquisition from successful tasks.
2026-02-17 03:24:38 -08:00
New platform adapter using signal-cli daemon (JSON-RPC HTTP + SSE). Requires Java runtime and phone number registration.
Update agent configuration for maximum tool-calling iterations - Increased the default maximum tool-calling iterations from 20 to 60 in the CLI configuration and related files, allowing for more complex tasks. - Updated documentation and comments to reflect the new recommended range for iterations, enhancing user guidance. - Implemented backward compatibility for loading max iterations from the root-level configuration, ensuring a smooth transition for existing users. - Adjusted the setup wizard to prompt for the maximum iterations setting, improving user experience during configuration.
2026-02-03 14:48:19 -08:00
Add todo tool for agent task planning and management Single `todo` tool that reads (no params) or writes (provide todos array with merge flag). In-memory TodoStore on AIAgent, no system prompt mutation, behavioral guidance in tool description only. State re-injected after context compression events. Gateway sessions hydrate from conversation history. Added to all platform toolsets. Also wired into RL agent_loop.py with per-run TodoStore and fixed browser_snapshot user_task passthrough from first user message.
2026-02-17 17:02:33 -08:00
**Reference:** OpenClaw has Signal support via signal-cli.
---
feat: update TODO.md with enhancements to skills and memory systems - Increased the tool count to 44+ and clarified the management of bundled and agent-managed skills. - Introduced a persistent memory system with MEMORY.md and USER.md for agent notes and user profiles. - Updated the storage evolution section to reflect the use of SQLite for sessions and clarified the organization of skills and memories. - Added current status of memory types implemented, highlighting progress in agent intelligence capabilities.
2026-02-19 18:47:45 -08:00
## 5. Plugin/Extension System 🔌
Add todo tool for agent task planning and management Single `todo` tool that reads (no params) or writes (provide todos array with merge flag). In-memory TodoStore on AIAgent, no system prompt mutation, behavioral guidance in tool description only. State re-injected after context compression events. Gateway sessions hydrate from conversation history. Added to all platform toolsets. Also wired into RL agent_loop.py with per-run TodoStore and fixed browser_snapshot user_task passthrough from first user message.
2026-02-17 17:02:33 -08:00
**Status:** Partially implemented (event hooks exist in `gateway/hooks.py`)
**Priority:** Medium
Full Python plugin interface that goes beyond the current hook system.
**What other agents do:**
- **OpenClaw**: Plugin SDK with tool-send capabilities, lifecycle phase hooks (before-agent-start, after-tool-call, model-override), plugin registry with install/uninstall.
- **Pi**: Extensions are TypeScript modules that can register tools, commands, keyboard shortcuts, custom UI widgets, overlays, status lines, dialogs, compaction hooks, raw terminal input listeners. Extremely comprehensive.
- **OpenCode**: MCP client support (stdio, SSE, StreamableHTTP), OAuth auth for MCP servers. Also has Copilot/Codex plugins.
- **Codex**: Full MCP integration with skill dependencies.
- **Cline**: MCP integration + lifecycle hooks with cancellation support.
**Our approach (phased):**
### Phase 1: Enhanced hooks
- Expand the existing `gateway/hooks.py` to support more events: `before-tool-call`, `after-tool-call`, `before-response`, `context-compress`, `session-end`
- Allow hooks to modify tool results (e.g., filter sensitive output)
### Phase 2: Plugin interface
- `~/.hermes/plugins/<name>/plugin.yaml` + `handler.py`
- Plugins can: register new tools, add CLI commands, subscribe to events, inject system prompt sections
- `hermes plugin list|install|uninstall|create` CLI commands
- Plugin discovery and validation on startup
fix(mcp): parallel discovery, user-visible logging, config validation - Discovery is now parallel (asyncio.gather) instead of sequential, fixing the 60s shared timeout issue with multiple servers - Startup messages use print() so users see connection status even with default log levels (the 'tools' logger is set to ERROR) - Summary line shows total tools and failed servers count - Validate conflicting config: warn if both 'url' and 'command' are present (HTTP takes precedence) - Update TODO.md: mark MCP as implemented, list remaining work - Add test for conflicting config detection (51 tests total) All 1163 tests pass.
2026-03-02 19:02:28 -08:00
### Phase 3: MCP support (industry standard) ✅ DONE
- ✅ MCP client that connects to external MCP servers (stdio + HTTP/StreamableHTTP)
- ✅ Config: `mcp_servers` in config.yaml with connection details
- ✅ Each MCP server's tools auto-registered as a dynamic toolset
- Future: Resources, Prompts, Progress notifications, `hermes mcp` CLI command
Add todo tool for agent task planning and management Single `todo` tool that reads (no params) or writes (provide todos array with merge flag). In-memory TodoStore on AIAgent, no system prompt mutation, behavioral guidance in tool description only. State re-injected after context compression events. Gateway sessions hydrate from conversation history. Added to all platform toolsets. Also wired into RL agent_loop.py with per-run TodoStore and fixed browser_snapshot user_task passthrough from first user message.
2026-02-17 17:02:33 -08:00
---
fix(mcp): parallel discovery, user-visible logging, config validation - Discovery is now parallel (asyncio.gather) instead of sequential, fixing the 60s shared timeout issue with multiple servers - Startup messages use print() so users see connection status even with default log levels (the 'tools' logger is set to ERROR) - Summary line shows total tools and failed servers count - Validate conflicting config: warn if both 'url' and 'command' are present (HTTP takes precedence) - Update TODO.md: mark MCP as implemented, list remaining work - Add test for conflicting config detection (51 tests total) All 1163 tests pass.
2026-03-02 19:02:28 -08:00
## 6. MCP (Model Context Protocol) Support 🔗 ✅ DONE
Add todo tool for agent task planning and management Single `todo` tool that reads (no params) or writes (provide todos array with merge flag). In-memory TodoStore on AIAgent, no system prompt mutation, behavioral guidance in tool description only. State re-injected after context compression events. Gateway sessions hydrate from conversation history. Added to all platform toolsets. Also wired into RL agent_loop.py with per-run TodoStore and fixed browser_snapshot user_task passthrough from first user message.
2026-02-17 17:02:33 -08:00
fix(mcp): parallel discovery, user-visible logging, config validation - Discovery is now parallel (asyncio.gather) instead of sequential, fixing the 60s shared timeout issue with multiple servers - Startup messages use print() so users see connection status even with default log levels (the 'tools' logger is set to ERROR) - Summary line shows total tools and failed servers count - Validate conflicting config: warn if both 'url' and 'command' are present (HTTP takes precedence) - Update TODO.md: mark MCP as implemented, list remaining work - Add test for conflicting config detection (51 tests total) All 1163 tests pass.
2026-03-02 19:02:28 -08:00
**Status:** Implemented (PR #301)
**Priority:** Complete
Add todo tool for agent task planning and management Single `todo` tool that reads (no params) or writes (provide todos array with merge flag). In-memory TodoStore on AIAgent, no system prompt mutation, behavioral guidance in tool description only. State re-injected after context compression events. Gateway sessions hydrate from conversation history. Added to all platform toolsets. Also wired into RL agent_loop.py with per-run TodoStore and fixed browser_snapshot user_task passthrough from first user message.
2026-02-17 17:02:33 -08:00
fix(mcp): parallel discovery, user-visible logging, config validation - Discovery is now parallel (asyncio.gather) instead of sequential, fixing the 60s shared timeout issue with multiple servers - Startup messages use print() so users see connection status even with default log levels (the 'tools' logger is set to ERROR) - Summary line shows total tools and failed servers count - Validate conflicting config: warn if both 'url' and 'command' are present (HTTP takes precedence) - Update TODO.md: mark MCP as implemented, list remaining work - Add test for conflicting config detection (51 tests total) All 1163 tests pass.
2026-03-02 19:02:28 -08:00
Native MCP client support with stdio and HTTP/StreamableHTTP transports, auto-discovery, reconnection with exponential backoff, env var filtering, and credential stripping. See `docs/mcp.md` for full documentation.
Add todo tool for agent task planning and management Single `todo` tool that reads (no params) or writes (provide todos array with merge flag). In-memory TodoStore on AIAgent, no system prompt mutation, behavioral guidance in tool description only. State re-injected after context compression events. Gateway sessions hydrate from conversation history. Added to all platform toolsets. Also wired into RL agent_loop.py with per-run TodoStore and fixed browser_snapshot user_task passthrough from first user message.
2026-02-17 17:02:33 -08:00
fix(mcp): parallel discovery, user-visible logging, config validation - Discovery is now parallel (asyncio.gather) instead of sequential, fixing the 60s shared timeout issue with multiple servers - Startup messages use print() so users see connection status even with default log levels (the 'tools' logger is set to ERROR) - Summary line shows total tools and failed servers count - Validate conflicting config: warn if both 'url' and 'command' are present (HTTP takes precedence) - Update TODO.md: mark MCP as implemented, list remaining work - Add test for conflicting config detection (51 tests total) All 1163 tests pass.
2026-03-02 19:02:28 -08:00
**Still TODO:**
- `hermes mcp` CLI subcommand (list/test/status)
- `hermes tools` UI integration for MCP toolsets
- MCP Resources and Prompts support
- OAuth authentication for remote servers
- Progress notifications for long-running tools
Add todo tool for agent task planning and management Single `todo` tool that reads (no params) or writes (provide todos array with merge flag). In-memory TodoStore on AIAgent, no system prompt mutation, behavioral guidance in tool description only. State re-injected after context compression events. Gateway sessions hydrate from conversation history. Added to all platform toolsets. Also wired into RL agent_loop.py with per-run TodoStore and fixed browser_snapshot user_task passthrough from first user message.
2026-02-17 17:02:33 -08:00
---
feat: update TODO.md with enhancements to skills and memory systems - Increased the tool count to 44+ and clarified the management of bundled and agent-managed skills. - Introduced a persistent memory system with MEMORY.md and USER.md for agent notes and user profiles. - Updated the storage evolution section to reflect the use of SQLite for sessions and clarified the organization of skills and memories. - Added current status of memory types implemented, highlighting progress in agent intelligence capabilities.
2026-02-19 18:47:45 -08:00
## 8. Filesystem Checkpointing / Rollback 🔄
Add todo tool for agent task planning and management Single `todo` tool that reads (no params) or writes (provide todos array with merge flag). In-memory TodoStore on AIAgent, no system prompt mutation, behavioral guidance in tool description only. State re-injected after context compression events. Gateway sessions hydrate from conversation history. Added to all platform toolsets. Also wired into RL agent_loop.py with per-run TodoStore and fixed browser_snapshot user_task passthrough from first user message.
2026-02-17 17:02:33 -08:00
**Status:** Not started
**Priority:** Low-Medium
feat: update TODO.md with enhancements to skills and memory systems - Increased the tool count to 44+ and clarified the management of bundled and agent-managed skills. - Introduced a persistent memory system with MEMORY.md and USER.md for agent notes and user profiles. - Updated the storage evolution section to reflect the use of SQLite for sessions and clarified the organization of skills and memories. - Added current status of memory types implemented, highlighting progress in agent intelligence capabilities.
2026-02-19 18:47:45 -08:00
Automatic filesystem snapshots after each agent loop iteration so the user can roll back destructive changes to their project.
Add todo tool for agent task planning and management Single `todo` tool that reads (no params) or writes (provide todos array with merge flag). In-memory TodoStore on AIAgent, no system prompt mutation, behavioral guidance in tool description only. State re-injected after context compression events. Gateway sessions hydrate from conversation history. Added to all platform toolsets. Also wired into RL agent_loop.py with per-run TodoStore and fixed browser_snapshot user_task passthrough from first user message.
2026-02-17 17:02:33 -08:00
**What other agents do:**
feat: update TODO.md with enhancements to skills and memory systems - Increased the tool count to 44+ and clarified the management of bundled and agent-managed skills. - Introduced a persistent memory system with MEMORY.md and USER.md for agent notes and user profiles. - Updated the storage evolution section to reflect the use of SQLite for sessions and clarified the organization of skills and memories. - Added current status of memory types implemented, highlighting progress in agent intelligence capabilities.
2026-02-19 18:47:45 -08:00
- **Cline**: Workspace checkpoints at each step with Compare/Restore UI
- **OpenCode**: Git-backed workspace snapshots per step, with weekly gc
- **Codex**: Sandboxed execution with commit-per-step, rollback on failure
Add todo tool for agent task planning and management Single `todo` tool that reads (no params) or writes (provide todos array with merge flag). In-memory TodoStore on AIAgent, no system prompt mutation, behavioral guidance in tool description only. State re-injected after context compression events. Gateway sessions hydrate from conversation history. Added to all platform toolsets. Also wired into RL agent_loop.py with per-run TodoStore and fixed browser_snapshot user_task passthrough from first user message.
2026-02-17 17:02:33 -08:00
**Our approach:**
feat: update TODO.md with enhancements to skills and memory systems - Increased the tool count to 44+ and clarified the management of bundled and agent-managed skills. - Introduced a persistent memory system with MEMORY.md and USER.md for agent notes and user profiles. - Updated the storage evolution section to reflect the use of SQLite for sessions and clarified the organization of skills and memories. - Added current status of memory types implemented, highlighting progress in agent intelligence capabilities.
2026-02-19 18:47:45 -08:00
- After each tool call (or batch of tool calls in a single turn) that modifies files, create a lightweight checkpoint of the affected files
- Git-based when the project is a repo: auto-commit to a detached/temporary branch (`hermes/checkpoints/<session>`) after each agent turn, squash or discard on session end
- Non-git fallback: tar snapshots of changed files in `~/.hermes/checkpoints/<session_id>/`
- `hermes rollback` CLI command to restore to a previous checkpoint
- Agent-accessible via a `checkpoint` tool: `list` (show available restore points), `restore` (roll back to a named point), `diff` (show what changed since a checkpoint)
- Configurable: off by default (opt-in via `config.yaml`), since auto-committing can be surprising
- Cleanup: checkpoints expire after session ends (or configurable retention period)
- Integration with the terminal backend: works with local, SSH, and Docker backends (snapshots happen on the execution host)
Add todo tool for agent task planning and management Single `todo` tool that reads (no params) or writes (provide todos array with merge flag). In-memory TodoStore on AIAgent, no system prompt mutation, behavioral guidance in tool description only. State re-injected after context compression events. Gateway sessions hydrate from conversation history. Added to all platform toolsets. Also wired into RL agent_loop.py with per-run TodoStore and fixed browser_snapshot user_task passthrough from first user message.
2026-02-17 17:02:33 -08:00
---
## Implementation Priority Order
Update agent configuration for maximum tool-calling iterations - Increased the default maximum tool-calling iterations from 20 to 60 in the CLI configuration and related files, allowing for more complex tasks. - Updated documentation and comments to reflect the new recommended range for iterations, enhancing user guidance. - Implemented backward compatibility for loading max iterations from the root-level configuration, ensuring a smooth transition for existing users. - Adjusted the setup wizard to prompt for the maximum iterations setting, improving user experience during configuration.
2026-02-03 14:48:19 -08:00
feat: update TODO.md with enhancements to skills and memory systems - Increased the tool count to 44+ and clarified the management of bundled and agent-managed skills. - Introduced a persistent memory system with MEMORY.md and USER.md for agent notes and user profiles. - Updated the storage evolution section to reflect the use of SQLite for sessions and clarified the organization of skills and memories. - Added current status of memory types implemented, highlighting progress in agent intelligence capabilities.
2026-02-19 18:47:45 -08:00
### Tier 1: Next Up
feat: add persistent memory system + SQLite session store Two-part implementation: Part A - Curated Bounded Memory: - New memory tool (tools/memory_tool.py) with MEMORY.md + USER.md stores - Character-limited (2200/1375 chars), § delimited entries - Frozen snapshot injected into system prompt at session start - Model manages pruning via replace/remove with substring matching - Usage indicator shown in system prompt header Part B - SQLite Session Store: - New hermes_state.py with SessionDB class, FTS5 full-text search - Gateway session.py rewritten to dual-write SQLite + legacy JSONL - Compression-triggered session splitting with parent_session_id chains - New session_search tool with Gemini Flash summarization of matched sessions - CLI session lifecycle (create on launch, close on exit) Also: - System prompt now cached per session, only rebuilt on compression (fixes prefix cache invalidation from date/time changes every turn) - Config version bumped to 3, hermes doctor checks for new artifacts - Disabled in batch_runner and RL environments
2026-02-19 00:57:31 -08:00
fix(mcp): parallel discovery, user-visible logging, config validation - Discovery is now parallel (asyncio.gather) instead of sequential, fixing the 60s shared timeout issue with multiple servers - Startup messages use print() so users see connection status even with default log levels (the 'tools' logger is set to ERROR) - Summary line shows total tools and failed servers count - Validate conflicting config: warn if both 'url' and 'command' are present (HTTP takes precedence) - Update TODO.md: mark MCP as implemented, list remaining work - Add test for conflicting config detection (51 tests total) All 1163 tests pass.
2026-03-02 19:02:28 -08:00
1. ~~MCP Support -- #6~~ ✅ Done (PR #301)
feat: add persistent memory system + SQLite session store Two-part implementation: Part A - Curated Bounded Memory: - New memory tool (tools/memory_tool.py) with MEMORY.md + USER.md stores - Character-limited (2200/1375 chars), § delimited entries - Frozen snapshot injected into system prompt at session start - Model manages pruning via replace/remove with substring matching - Usage indicator shown in system prompt header Part B - SQLite Session Store: - New hermes_state.py with SessionDB class, FTS5 full-text search - Gateway session.py rewritten to dual-write SQLite + legacy JSONL - Compression-triggered session splitting with parent_session_id chains - New session_search tool with Gemini Flash summarization of matched sessions - CLI session lifecycle (create on launch, close on exit) Also: - System prompt now cached per session, only rebuilt on compression (fixes prefix cache invalidation from date/time changes every turn) - Config version bumped to 3, hermes doctor checks for new artifacts - Disabled in batch_runner and RL environments
2026-02-19 00:57:31 -08:00
feat: implement code execution sandbox for programmatic tool calling - Introduced a new `execute_code` tool that allows the agent to run Python scripts that call Hermes tools via RPC, reducing the number of round trips required for tool interactions. - Added configuration options for timeout and maximum tool calls in the sandbox environment. - Updated the toolset definitions to include the new code execution capabilities, ensuring integration across platforms. - Implemented comprehensive tests for the code execution sandbox, covering various scenarios including tool call limits and error handling. - Enhanced the CLI and documentation to reflect the new functionality, providing users with clear guidance on using the code execution tool.
2026-02-19 23:23:43 -08:00
### Tier 2: Quality of Life
feat: add persistent memory system + SQLite session store Two-part implementation: Part A - Curated Bounded Memory: - New memory tool (tools/memory_tool.py) with MEMORY.md + USER.md stores - Character-limited (2200/1375 chars), § delimited entries - Frozen snapshot injected into system prompt at session start - Model manages pruning via replace/remove with substring matching - Usage indicator shown in system prompt header Part B - SQLite Session Store: - New hermes_state.py with SessionDB class, FTS5 full-text search - Gateway session.py rewritten to dual-write SQLite + legacy JSONL - Compression-triggered session splitting with parent_session_id chains - New session_search tool with Gemini Flash summarization of matched sessions - CLI session lifecycle (create on launch, close on exit) Also: - System prompt now cached per session, only rebuilt on compression (fixes prefix cache invalidation from date/time changes every turn) - Config version bumped to 3, hermes doctor checks for new artifacts - Disabled in batch_runner and RL environments
2026-02-19 00:57:31 -08:00
feat: implement code execution sandbox for programmatic tool calling - Introduced a new `execute_code` tool that allows the agent to run Python scripts that call Hermes tools via RPC, reducing the number of round trips required for tool interactions. - Added configuration options for timeout and maximum tool calls in the sandbox environment. - Updated the toolset definitions to include the new code execution capabilities, ensuring integration across platforms. - Implemented comprehensive tests for the code execution sandbox, covering various scenarios including tool call limits and error handling. - Enhanced the CLI and documentation to reflect the new functionality, providing users with clear guidance on using the code execution tool.
2026-02-19 23:23:43 -08:00
3. Local Browser Control via CDP -- #3
4. Plugin/Extension System -- #5
feat: add persistent memory system + SQLite session store Two-part implementation: Part A - Curated Bounded Memory: - New memory tool (tools/memory_tool.py) with MEMORY.md + USER.md stores - Character-limited (2200/1375 chars), § delimited entries - Frozen snapshot injected into system prompt at session start - Model manages pruning via replace/remove with substring matching - Usage indicator shown in system prompt header Part B - SQLite Session Store: - New hermes_state.py with SessionDB class, FTS5 full-text search - Gateway session.py rewritten to dual-write SQLite + legacy JSONL - Compression-triggered session splitting with parent_session_id chains - New session_search tool with Gemini Flash summarization of matched sessions - CLI session lifecycle (create on launch, close on exit) Also: - System prompt now cached per session, only rebuilt on compression (fixes prefix cache invalidation from date/time changes every turn) - Config version bumped to 3, hermes doctor checks for new artifacts - Disabled in batch_runner and RL environments
2026-02-19 00:57:31 -08:00
feat: implement code execution sandbox for programmatic tool calling - Introduced a new `execute_code` tool that allows the agent to run Python scripts that call Hermes tools via RPC, reducing the number of round trips required for tool interactions. - Added configuration options for timeout and maximum tool calls in the sandbox environment. - Updated the toolset definitions to include the new code execution capabilities, ensuring integration across platforms. - Implemented comprehensive tests for the code execution sandbox, covering various scenarios including tool call limits and error handling. - Enhanced the CLI and documentation to reflect the new functionality, providing users with clear guidance on using the code execution tool.
2026-02-19 23:23:43 -08:00
### Tier 3: Nice to Have
feat: add persistent memory system + SQLite session store Two-part implementation: Part A - Curated Bounded Memory: - New memory tool (tools/memory_tool.py) with MEMORY.md + USER.md stores - Character-limited (2200/1375 chars), § delimited entries - Frozen snapshot injected into system prompt at session start - Model manages pruning via replace/remove with substring matching - Usage indicator shown in system prompt header Part B - SQLite Session Store: - New hermes_state.py with SessionDB class, FTS5 full-text search - Gateway session.py rewritten to dual-write SQLite + legacy JSONL - Compression-triggered session splitting with parent_session_id chains - New session_search tool with Gemini Flash summarization of matched sessions - CLI session lifecycle (create on launch, close on exit) Also: - System prompt now cached per session, only rebuilt on compression (fixes prefix cache invalidation from date/time changes every turn) - Config version bumped to 3, hermes doctor checks for new artifacts - Disabled in batch_runner and RL environments
2026-02-19 00:57:31 -08:00
feat: implement code execution sandbox for programmatic tool calling - Introduced a new `execute_code` tool that allows the agent to run Python scripts that call Hermes tools via RPC, reducing the number of round trips required for tool interactions. - Added configuration options for timeout and maximum tool calls in the sandbox environment. - Updated the toolset definitions to include the new code execution capabilities, ensuring integration across platforms. - Implemented comprehensive tests for the code execution sandbox, covering various scenarios including tool call limits and error handling. - Enhanced the CLI and documentation to reflect the new functionality, providing users with clear guidance on using the code execution tool.
2026-02-19 23:23:43 -08:00
5. Session Branching / Checkpoints -- #7
6. Filesystem Checkpointing / Rollback -- #8
7. Signal Integration -- #4
Reference in New Issue Copy Permalink