* feat: add /branch (/fork) command for session branching
Inspired by Claude Code's /branch command. Creates a copy of the current
session's conversation history in a new session, allowing the user to
explore a different approach without losing the original.
Works like 'git checkout -b' for conversations:
- /branch — auto-generates a title from the parent session
- /branch my-idea — uses a custom title
- /fork — alias for /branch
Implementation:
- CLI: _handle_branch_command() in cli.py
- Gateway: _handle_branch_command() in gateway/run.py
- CommandDef with 'fork' alias in commands.py
- Uses existing parent_session_id field in session DB
- Uses get_next_title_in_lineage() for auto-numbered branches
- 14 tests covering session creation, history copy, parent links,
title generation, edge cases, and agent sync
* fix: clear ghost status-bar lines on terminal resize
When the terminal shrinks (e.g. un-maximize), the emulator reflows
previously full-width rows (status bar, input rules) into multiple
narrower rows. prompt_toolkit's _on_resize only cursor_up()s by the
stored layout height, missing the extra rows from reflow — leaving
ghost duplicates of the status bar visible.
Fix: monkey-patch Application._on_resize to detect width shrinks,
calculate the extra rows created by reflow, and inflate the renderer's
cursor_pos.y so the erase moves up far enough to clear ghosts.
Two pre-existing issues causing test_file_read_guards timeouts on CI:
1. agent/redact.py: _ENV_ASSIGN_RE used unbounded [A-Z_]* with
IGNORECASE, matching any letter/underscore to end-of-string at
each position → O(n²) backtracking on 100K+ char inputs.
Bounded to {0,50} since env var names are never that long.
2. tools/file_tools.py: redact_sensitive_text() ran BEFORE the
character-count guard, so oversized content (that would be rejected
anyway) went through the expensive regex first. Reordered to check
size limit before redaction.
- HERMES_AGENT_TIMEOUT=0 now means no limit (infinite execution)
- Periodic 'still working' notifications every 10 minutes for long tasks
- Timeout error message now tells users how to increase the limit
- Stale-lock eviction handles infinite timeout correctly (float inf TTL)
Clearer workflow with validation/chunking steps, expanded description
with trigger terms for better agent matching, tightened error handling.
Fixed stray pipe character in original PR diff.
Based on PR #4778 by fernandezbaptiste.
Co-authored-by: fernandezbaptiste <fernandezbaptiste@users.noreply.github.com>
The config key skills.external_dirs and core resolution (get_all_skills_dirs,
get_external_skills_dirs in agent/skill_utils.py) already existed but several
code paths still only scanned SKILLS_DIR. Now external dirs are respected
everywhere:
- skills_categories(): scan all dirs for category discovery
- _get_category_from_path(): resolve categories against any skills root
- skill_manager_tool._find_skill(): search all dirs for edit/patch/delete
- credential_files.get_skills_directory_mount(): mount all dirs into
Docker/Singularity containers (external dirs at external_skills/<idx>)
- credential_files.iter_skills_files(): list files from all dirs for
Modal/Daytona upload
- tools/environments/ssh.py: rsync all skill dirs to remote hosts
- gateway _check_unavailable_skill(): check disabled skills across all dirs
Usage in config.yaml:
skills:
external_dirs:
- ~/repos/agent-skills/hermes
- /shared/team-skills
When a skill is active, user_message contains the full SKILL.md content
injected by the skill system. This bloated string was being passed to
memory provider sync_all(), queue_prefetch_all(), and prefetch_all(),
causing providers with query size limits (e.g. Honcho's 10K char limit)
to fail.
Both call sites now use original_user_message (the clean user input,
already defined at line 6516) instead of the skill-inflated user_message:
- Pre-turn prefetch (line ~6695): prefetch_all() query
- Post-turn sync (line ~8672): sync_all() + queue_prefetch_all()
Fixes#4889
The original filter (if v) silently drops False and 0, so
'rerank: false' in mem0.json would be ignored. Use explicit
None/empty-string check to preserve intentional falsy values.
When mem0.json exists but is missing the api_key (e.g. after running
`hermes memory setup`), the plugin reports "not available" even though
MEM0_API_KEY is set in .env. This happens because _load_config()
returns the JSON file contents verbatim, never falling back to env vars.
Use env vars as the base config and let mem0.json override individual
keys on top, so both config sources work together.
Fixes: mem0 plugin shows "not available" despite valid MEM0_API_KEY in .env
OpenViking is multi-tenant and requires X-OpenViking-Account and
X-OpenViking-User headers. Without them, API calls like POST
/api/v1/search/find fail on authenticated servers.
Add both headers to _VikingClient._headers(), read from env vars
OPENVIKING_ACCOUNT (default: root) and OPENVIKING_USER (default:
default). All instantiation sites inherit the fix automatically.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The tool_preview_length: 0 (unlimited) config change from e314833c
removed truncation from gateway progress messages in all/new modes.
This caused full terminal commands, code blocks, and file paths to
appear as permanent messages in Telegram -- the old 40-char truncation
was the correct behavior for messaging platforms.
Now:
- all/new modes: always truncate previews to 40 chars (old behavior)
- verbose mode: respects tool_preview_length config for JSON args cap
Reported by Paulclgro and socialsurfer on Discord.
Input like /Users/ironin/file.md:45-46 was routed to process_command()
because it starts with /. Added _looks_like_slash_command() which checks
whether the first word contains additional / characters — commands never
do (/help, /model), paths always do (/Users/foo/bar.md).
Applied to both process_loop routing and handle_enter interrupt bypass.
Preserves prefix matching (/h → /help) since short prefixes still pass
the check.
Based on PR #4782 by iRonin.
Co-authored-by: iRonin <iRonin@users.noreply.github.com>
The base adapter's active-session guard queues all messages when an agent
is running. This creates a deadlock for /approve and /deny: the agent
thread is blocked on threading.Event.wait() in tools/approval.py waiting
for resolve_gateway_approval(), but the /approve command is queued waiting
for the agent to finish.
Dispatch /approve and /deny directly to the message handler (which routes
to gateway/run.py's _handle_approve_command) without going through
_process_message_background — avoids spawning a competing background task
that would mess with session lifecycle/guards.
Fixes#4898
Co-authored-by: mechovation (original diagnosis in PR #4904)
* fix(gateway): add message deduplication to Discord and Slack adapters (#4777)
Discord RESUME replays events after reconnects (~7/day observed),
and Slack Socket Mode can redeliver events if the ack was lost.
Neither adapter tracked which messages were already processed,
causing duplicate bot responses.
Add _seen_messages dedup cache (message ID → timestamp) with 5-min
TTL and 2000-entry cap to both adapters, matching the pattern already
used by Mattermost, Matrix, WeCom, Feishu, DingTalk, and Email.
The check goes at the very top of the message handler, before any
other logic, so replayed events are silently dropped.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: prevent duplicate messages on partial stream delivery
When streaming fails after tokens are already delivered to the platform,
_interruptible_streaming_api_call re-raised the error into the outer
retry loop, which would make a new API call — creating a duplicate
message.
Now checks deltas_were_sent before re-raising: if partial content was
already streamed, returns a stub response instead. The outer loop treats
the turn as complete (no retry, no fallback, no duplicate).
Inspired by PR #4871 (@trevorgordon981) which identified the bug.
This implementation avoids monkey-patching exception objects and keeps
the fix within the streaming call boundary.
---------
Co-authored-by: Mibayy <mibayy@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The Anthropic SDK appends /v1/messages to the base_url, so OpenCode's
base URL https://opencode.ai/zen/go/v1 produced a double /v1 path
(https://opencode.ai/zen/go/v1/v1/messages), causing 404s for MiniMax
models. Strip trailing /v1 when api_mode is anthropic_messages.
Also adds MiMo-V2-Pro, MiMo-V2-Omni, and MiniMax-M2.5 to the OpenCode
Go model lists per their updated docs.
Fixes#4890
- 7 new tests covering skill binding, fallthrough, coercion
- Docs section in telegram.md with config format, field reference,
comparison table, and thread_id discovery tip
Reads config.extra['group_topics'] to bind skills to specific thread_ids
in supergroup/forum chats. Mirrors the dm_topics skill injection pattern
but for group chat_type. Enables per-topic skill auto-loading in Falcon HQ.
Config format:
platforms.telegram.extra.group_topics:
- chat_id: -1003853746818
topics:
- name: FalconConnect
thread_id: 5
skill: falconconnect-architecture
- Add contextvars.Token[str] type hints to set/reset_current_session_key
- Use get_current_session_key(default='') in terminal_tool.py for background
process session tracking, fixing the same env var race for concurrent
gateway sessions spawning background processes
The '📨 Queued:' label was misleading — it looked like the message was
silently deferred when it was actually being sent immediately after the
interrupt. Changed to '⚡ Sending after interrupt:' with multi-message
count when the user typed several messages during agent execution.
Added comment documenting that this code path only applies when
busy_input_mode == 'interrupt' (the default).
Based on PR #4821 by iRonin.
Co-authored-by: iRonin <iRonin@users.noreply.github.com>
The original PR missed the third CLI restore path — the /session switch
command that loads history via get_messages_as_conversation() without
stripping session_meta entries.
Add a provider-agnostic role allowlist guard to _sanitize_api_messages()
that drops messages with roles not accepted by the chat-completions API
(e.g. session_meta). This prevents CLI resume/session restore from
leaking transcript-only metadata into the outgoing messages payload.
Two layers of defense:
1. API-boundary guard: _sanitize_api_messages() now filters messages by
role allowlist (system/user/assistant/tool/function/developer) before
the existing orphaned tool-call repair logic. This protects all
current and future call paths.
2. CLI restore defense-in-depth: Both session restore paths in cli.py
now strip session_meta entries before loading history into
conversation_history, matching the existing gateway behavior.
Closes#4715
gateway stop and restart previously called kill_gateway_processes() which
scans ps aux and kills ALL gateway processes across all profiles. Starting
a profile gateway would nuke the main one (and vice versa).
Now:
- hermes gateway stop → only kills the current profile's gateway (PID file)
- hermes -p work gateway stop → only kills the 'work' profile's gateway
- hermes gateway stop --all → kills every gateway process (old behavior)
- hermes gateway restart → profile-scoped for manual fallback path
- hermes update → discovers and restarts ALL profile gateways (systemctl
list-units hermes-gateway*) since the code update is shared
Added stop_profile_gateway() which uses the HERMES_HOME-scoped PID file
instead of global process scanning.
Windows (CRLF) and old Mac (CR) line endings are normalised to LF
before the 5-line collapse threshold is checked in handle_paste.
Without this, markdown copied from Windows sources contains \r\n but
the line counter (pasted_text.count('\n')) still works — however
buf.insert_text() leaves bare \r characters in the buffer which some
terminals render by moving the cursor to the start of the line,
making multi-line pastes appear as a single overwritten line.
- Add .zip to SUPPORTED_DOCUMENT_TYPES so gateway platforms (Telegram,
Slack, Discord) cache uploaded zip files instead of rejecting them.
- Add get_cache_directory_mounts() and iter_cache_files() to
credential_files.py for host-side cache directory passthrough
(documents, images, audio, screenshots).
- Docker: bind-mount cache dirs read-only alongside credentials/skills.
Changes are live (bind mount semantics).
- Modal: mount cache files at sandbox creation + resync before each
command via _sync_files() with mtime+size change detection.
- Handles backward-compat with legacy dir names (document_cache,
image_cache, audio_cache, browser_screenshots) via get_hermes_dir().
- Container paths always use the new cache/<subdir> layout regardless
of host layout.
This replaces the need for a dedicated extract_archive tool (PR #4819)
— the agent can now use standard terminal commands (unzip, tar) on
uploaded files inside remote containers.
Closes: related to PR #4819 by kshitijk4poor
Three fixes for memory+profile isolation bugs:
1. memory_tool.py: Replace module-level MEMORY_DIR constant with
get_memory_dir() function that calls get_hermes_home() dynamically.
The old constant was cached at import time and could go stale if
HERMES_HOME changed after import. Internal MemoryStore methods now
call get_memory_dir() directly. MEMORY_DIR kept as backward-compat
alias.
2. profiles.py: profile create --clone now copies MEMORY.md and USER.md
from the source profile. These curated memory files are part of the
agent's identity (same as SOUL.md) and should carry over on clone.
3. holographic plugin: initialize() now expands $HERMES_HOME and
${HERMES_HOME} in the db_path config value, so users can write
'db_path: $HERMES_HOME/memory_store.db' and it resolves to the
active profile directory, not the default home.
Tests updated to mock get_memory_dir() alongside the legacy MEMORY_DIR.
uv.lock was stale at v0.5.0 and missing exa-py (core dep), causing
ModuleNotFoundError for Nix flake builds. Also syncs faster-whisper
placement (core → voice extra), adds feishu/debugpy/lark-oapi extras.
Fixes#4648
Credit to @lvnilesh for identifying the issue in PR #4649.
The pipeline prefetch design was firing \`brv query\` in a background
thread *after* each response, meaning the context injected at turn N
was from turn N-1's message — and the first turn got no BRV context
at all. Replace the async prefetch pipeline with a synchronous query
in \`prefetch()\` so recall runs before the first API call on every
turn. Make \`queue_prefetch()\` a no-op and remove the now-unused
pipeline state.
Memory provider tools (hindsight_retain, honcho_search, etc.) were
advertised to the model via tool schemas but failed with 'Unknown tool'
at execution time. The concurrent path (_invoke_tool) correctly checks
self._memory_manager.has_tool() before falling through to the registry,
but the sequential path (_execute_tool_calls_sequential) was never
updated with this check. Since sequential is the default for single
tool calls, memory provider tools always hit the registry dispatcher
which returns 'Unknown tool' because they're not registered there.
Add the memory_manager dispatch check between the delegate_task handler
and the quiet_mode fallthrough in the sequential path, with proper
spinner/display handling to match the existing pattern.
Reported by KiBenderOP — all memory providers affected (Honcho,
Hindsight, Holographic, etc.).
The API server adapter created AIAgent instances without passing
session_db, so conversations via Open WebUI and other OpenAI-compatible
frontends were never persisted to state.db. This meant 'hermes sessions
list' showed no API server sessions — they were effectively stateless.
Changes:
- Add _ensure_session_db() helper for lazy SessionDB initialization
- Pass session_db=self._ensure_session_db() in _create_agent()
- Refactor existing X-Hermes-Session-Id handler to use the shared helper
Sessions now persist with source='api_server' and are visible alongside
CLI and gateway sessions in hermes sessions list/search.
Two fixes for Discord exec approval:
1. Register /approve and /deny as native Discord slash commands so they
appear in Discord's command picker (autocomplete). Previously they
were only handled as text commands, so users saw 'no commands found'
when typing /approve.
2. Wire up the existing ExecApprovalView button UI (was dead code):
- ExecApprovalView now calls resolve_gateway_approval() to actually
unblock the waiting agent thread when a button is clicked
- Gateway's _approval_notify_sync() detects adapters with
send_exec_approval() and routes through the button UI
- Added 'Allow Session' button for parity with /approve session
- send_exec_approval() now accepts session_key and metadata for
thread support
- Graceful fallback to text-based /approve prompt if button send fails
Also updates test mocks to include grey/secondary ButtonStyle and
purple Color (used by new button styles).
Three interconnected bugs caused `hermes skills config` per-platform
settings to be silently ignored:
1. telegram_menu_commands() never filtered disabled skills — all skills
consumed menu slots regardless of platform config, hitting Telegram's
100 command cap. Now loads disabled skills for 'telegram' and excludes
them from the menu.
2. Gateway skill dispatch executed disabled skills because
get_skill_commands() (process-global cache) only filters by the global
disabled list at scan time. Added per-platform check before execution,
returning an actionable 'skill is disabled' message.
3. get_disabled_skill_names() only checked HERMES_PLATFORM env var, but
the gateway sets HERMES_SESSION_PLATFORM instead. Added
HERMES_SESSION_PLATFORM as fallback, plus an explicit platform=
parameter for callers that know their platform (menu builder, gateway
dispatch). Also added platform to prompt_builder's skills cache key
so multi-platform gateways get correct per-platform skill prompts.
Reported by SteveSkedasticity (CLAW community).
When the agent is blocked on a dangerous command approval (threading.Event
wait inside tools/approval.py), incoming /approve and /deny commands were
falling through to the generic interrupt path instead of being dispatched
to their command handlers. The interrupt sets _interrupt_requested on the
agent, but the agent thread is blocked on event.wait() — not checking the
flag. Result: approval times out after 300s (5 minutes) before executing.
Fix: intercept /approve and /deny in the running-agent early-intercept
block (alongside /stop, /new, /queue) and route directly to
_handle_approve_command / _handle_deny_command.
Addresses common questions from the Nous Research community Discord:
- Multi-model workflows via delegation config
- WhatsApp per-chat binding limitations and workarounds
- Controlling tool progress display on Telegram
- Per-platform skills config and Telegram 100-command limit
- Shared thread sessions across multiple users
- Exporting/migrating Hermes to a new machine
- Permission denied on shell reload after install
- HTTP 400 on first agent run
When config.yaml has 'mcp_servers:' with no value, YAML parses it as
None. dict.get('mcp_servers', {}) only returns the default when the key
is absent, not when it's explicitly None. Use 'or {}' pattern to handle
both cases, matching the other two assignment sites in the same file.
Four fixes for MCP server stability issues reported by community member
(terminal lockup, zombie processes, escape sequence pollution, startup hang):
1. MCP reload timeout guard (cli.py): _check_config_mcp_changes now runs
_reload_mcp in a separate daemon thread with a 30s hard timeout. Previously,
a hung MCP server could block the process_loop thread indefinitely, freezing
the entire TUI (user can type but nothing happens, only Ctrl+D/Ctrl+\ work).
2. MCP stdio subprocess PID tracking (mcp_tool.py): Tracks child PIDs spawned
by stdio_client via before/after snapshots of /proc children. On shutdown,
_stop_mcp_loop force-kills any tracked PIDs that survived the SDK's graceful
SIGTERM→SIGKILL cleanup. Prevents zombie MCP server processes from
accumulating across sessions.
3. MCP event loop exception handler (mcp_tool.py): Installs
_mcp_loop_exception_handler on the MCP background event loop — same pattern
as the existing _suppress_closed_loop_errors on prompt_toolkit's loop.
Suppresses benign 'Event loop is closed' RuntimeError from httpx transport
__del__ during MCP shutdown. Salvaged from PR #2538 (acsezen).
4. MCP OAuth non-blocking (mcp_oauth.py): Replaces blocking input() call in
_wait_for_callback with OAuthNonInteractiveError raise. Adds _is_interactive()
TTY detection. In non-interactive environments, build_oauth_auth() still
returns a provider (cached tokens + refresh work), but the callback handler
raises immediately instead of blocking the MCP event loop for 120s. Re-raises
OAuth setup failures in _run_http so failed servers are reported cleanly
without blocking others. Salvaged from PRs #4521 (voidborne-d) and #4465
(heathley).
Closes#2537, closes#4462
Related: #4128, #3436
Three fixes for long-running gateway sessions that enter a death spiral
when API disconnects prevent token data collection, which prevents
compression, which causes more disconnects:
Layer 1 — Stale token counter fallback (run_agent.py in-loop):
When last_prompt_tokens is 0 (stale after API disconnect or provider
returned no usage data), fall back to estimate_messages_tokens_rough()
instead of passing 0 to should_compress(), which would never fire.
Layer 2 — Server disconnect heuristic (run_agent.py error handler):
When ReadError/RemoteProtocolError hits a large session (>60% context
or >200 messages), treat it as a context-length error and trigger
compression rather than burning through retries that all fail the
same way.
Layer 3 — Hard message count limit (gateway/run.py hygiene):
Force compression when a session exceeds 400 messages, regardless of
token estimates. This catches runaway growth even when all token-based
checks fail due to missing API data.
Based on the analysis from PR #2157 by ygd58 — the gateway threshold
direction fix (1.4x multiplier) was already resolved on main.
Anthropic returns HTTP 429 'Extra usage is required for long context
requests' when a Claude Max subscription doesn't include the 1M context
tier. This is NOT a transient rate limit — retrying won't help.
Only applies to Sonnet models (Opus 1M is general access). Detects
this specific error before the generic rate-limit handler and:
1. Reduces context_length from 1M to 200k (the standard tier)
2. Triggers context compression to fit
3. Retries with the reduced context
The reduction is session-scoped (not persisted) so it auto-recovers
if the user later enables extra usage on their subscription.
Fixes: Sonnet 4.6 instant rate limits on Claude Max without extra usage
Anthropic returns HTTP 429 'Extra usage is required for long context
requests' when a Claude Max subscription doesn't include the 1M context
tier. This is NOT a transient rate limit — retrying won't help.
Detect this specific error before the generic rate-limit handler and:
1. Reduce context_length from 1M to 200k (the standard tier)
2. Trigger context compression to fit
3. Retry with the reduced context
The reduction is session-scoped (not persisted) so it auto-recovers
if the user later enables extra usage on their subscription.
Fixes: Sonnet 4.6 instant rate limits on Claude Max without extra usage
