format_token_count_compact() used unconditional rstrip("0") to clean up
decimal trailing zeros (e.g. "1.50" → "1.5"), but this also stripped
meaningful trailing zeros from whole numbers ("260" → "26", "100" → "1").
Guard the strip behind a decimal-point check.
Co-authored-by: kshitijk4poor <82637225+kshitijk4poor@users.noreply.github.com>
When the CLI is active, sys.stdout is prompt_toolkit's StdoutProxy which
queues writes and injects newlines around each flush(). This causes every
\r spinner frame to land on its own line instead of overwriting the
previous one, producing visible flickering where the spinner and status
bar repeatedly swap positions.
The CLI already renders spinner state via a dedicated TUI widget
(_spinner_text / get_spinner_text), so KawaiiSpinner's \r-based loop is
redundant under StdoutProxy. Detect the proxy and suppress the animation
entirely — the thread still runs to preserve start()/stop() semantics.
Also removes the 0.4s flush rate-limit workaround that was papering over
the same issue, and cleans up the unused _last_flush_time attribute.
Salvaged from PR #2908 by Mibayy (fixed _raw -> raw detection, dropped
unrelated bundled changes).
build_skills_system_prompt() was calling _read_skill_conditions() which
re-read each SKILL.md file to extract conditional activation fields.
The frontmatter was already parsed by _parse_skill_file() earlier in
the same loop. Extract conditions inline from the existing frontmatter
dict instead, saving one file read per skill (~80+ on a typical setup).
Salvaged from PR #2827 by InB4DevOps.
- threshold: 0.80 → 0.50 (compress at 50%, not 80%)
- target_ratio: 0.40 → 0.20, now relative to threshold not total context
(20% of 50% = 10% of context as tail budget)
- summary ceiling: 32K → 12K (Gemini can't output more than ~12K)
- Updated DEFAULT_CONFIG, config display, example config, and tests
The summary_target_tokens parameter was accepted in the constructor,
stored on the instance, and never used — the summary budget was always
computed from hardcoded module constants (_SUMMARY_RATIO=0.20,
_MAX_SUMMARY_TOKENS=8000). This caused two compounding problems:
1. The config value was silently ignored, giving users no control
over post-compression size.
2. Fixed budgets (20K tail, 8K summary cap) didn't scale with
context window size. Switching from a 1M-context model to a
200K model would trigger compression that nuked 350K tokens
of conversation history down to ~30K.
Changes:
- Replace summary_target_tokens with summary_target_ratio (default 0.40)
which sets the post-compression target as a fraction of context_length.
Tail token budget and summary cap now scale proportionally:
MiniMax 200K → ~80K post-compression
GPT-5 1M → ~400K post-compression
- Change threshold_percent default: 0.50 → 0.80 (don't fire until
80% of context is consumed)
- Change protect_last_n default: 4 → 20 (preserve ~10 full turns)
- Summary token cap scales to 5% of context (was fixed 8K), capped
at 32K ceiling
- Read target_ratio and protect_last_n from config.yaml compression
section (both are now configurable)
- Remove hardcoded summary_target_tokens=500 from run_agent.py
- Add 5 new tests for ratio scaling, clamping, and new defaults
Move OpenRouter to position 1 in the setup wizard's provider list
to match hermes model ordering. Update default selection index and
fix test expectations for the new ordering.
Setup order: OpenRouter → Nous Portal → Codex → Custom → ...
When AsyncOpenAI clients are garbage-collected after the event loop
closes, their AsyncHttpxClientWrapper.__del__ tries to schedule
aclose() on the dead loop, causing RuntimeError: Event loop is closed.
prompt_toolkit catches this as an unhandled exception and shows
'Press ENTER to continue...' which blocks CLI exit.
Fix: Add shutdown_cached_clients() to auxiliary_client.py that marks
all cached async clients' underlying httpx transport as CLOSED before
GC runs. This prevents __del__ from attempting the aclose() call.
- _force_close_async_httpx(): sets httpx AsyncClient._state to CLOSED
- shutdown_cached_clients(): iterates _client_cache, closes sync clients
normally and marks async clients as closed
- Also fix stale client eviction in _get_cached_client to mark evicted
async clients as closed (was just del-ing them, triggering __del__)
- Call shutdown_cached_clients() from _run_cleanup() in cli.py
The context length resolver was querying the /models endpoint for known
providers like GitHub Copilot, which returns a provider-imposed limit
(128k) instead of the model's actual context window (400k for gpt-5.4).
Since this check happened before the models.dev lookup, the wrong value
won every time.
Fix:
- Add api.githubcopilot.com and models.github.ai to _URL_TO_PROVIDER
- Skip the endpoint metadata probe for known providers — their /models
data is unreliable for context length. models.dev has the correct
per-provider values.
Reported by danny [DUMB] — gpt-5.4 via Copilot was resolving to 128k
instead of the correct 400k from models.dev.
When a non-OpenRouter provider (e.g. minimax, anthropic) is set in
config.yaml but its API key is missing, Hermes silently fell back to
OpenRouter, causing confusing 404 errors.
Now checks if the user explicitly configured a provider before falling
back. Explicit providers raise RuntimeError with a clear message naming
the missing env var. Auto/openrouter/custom providers still fall through
to OpenRouter as before.
Three code paths fixed:
- run_agent.py AIAgent.__init__ — main client initialization
- auxiliary_client.py call_llm — sync auxiliary calls
- auxiliary_client.py call_llm_streaming — async auxiliary calls
Based on PR #2272 by @StefanIsMe. Applied manually to fix a
pconfig NameError in the original and extend to call_llm_streaming.
Co-authored-by: StefanIsMe <StefanIsMe@users.noreply.github.com>
Recent versions of llama.cpp moved the server properties endpoint from
/props to /v1/props (consistent with the /v1 API prefix convention).
The server-type detection path and the n_ctx reading path both used the
old /props URL, which returns 404 on current builds. This caused the
allocated context window size to fall back to a hardcoded default,
resulting in an incorrect (too small) value being displayed in the TUI
context bar.
Fix: try /v1/props first, fall back to /props for backward compatibility
with older llama.cpp builds. Both paths are now handled gracefully.
Two bugs in the auxiliary provider auto-detection chain:
1. Expired Codex JWT blocks the auto chain: _read_codex_access_token()
returned any stored token without checking expiry, preventing fallback
to working providers. Now decodes JWT exp claim and returns None for
expired tokens.
2. Auxiliary Anthropic client missing OAuth identity transforms:
_AnthropicCompletionsAdapter always called build_anthropic_kwargs with
is_oauth=False, causing 400 errors for OAuth tokens. Now detects OAuth
tokens via _is_oauth_token() and propagates the flag through the
adapter chain.
Cherry-picked from PR #2378 by 0xbyt4. Fixed test_api_key_no_oauth_flag
to mock resolve_anthropic_token directly (env var alone was insufficient).
redact_sensitive_text() now returns early for None and coerces other
non-string values to str before applying regex-based redaction,
preventing TypeErrors in logging/tool-output paths.
Cherry-picked from PR #2369 by aydnOktay.
On the native Anthropic Messages API path, convert_messages_to_anthropic()
moves top-level cache_control on role:tool messages inside the tool_result
block. On OpenRouter (chat_completions), no such conversion happens — the
unexpected top-level field causes a silent hang on the second tool call.
Add native_anthropic parameter to _apply_cache_marker() and
apply_anthropic_cache_control(). When False (OpenRouter), role:tool messages
are skipped entirely. When True (native Anthropic), existing behaviour is
preserved.
Fixes#2362
Only honor config.model.base_url for Anthropic resolution when
config.model.provider is actually "anthropic". This prevents a Codex
(or other provider) base_url from leaking into Anthropic runtime and
auxiliary client paths, which would send requests to the wrong
endpoint.
Closes#2384
Add @file:path, @folder:dir, @diff, @staged, @git:N, and @url:
references that expand inline before the message reaches the LLM.
Supports line ranges (@file:main.py:10-50), token budget enforcement
(soft warn at 25%, hard block at 50%), and path sandboxing for gateway.
Core module from PR #2090 by @kshitijk4poor. CLI and gateway wiring
rewritten against current main. Fixed asyncio.run() crash when called
from inside a running event loop (gateway).
Closes#682.
Two fixes for local model context detection:
1. Hardcoded DEFAULT_CONTEXT_LENGTHS matching was case-sensitive.
'qwen' didn't match 'Qwen3.5-9B-Q4_K_M.gguf' because of the
capital Q. Now uses model.lower() for comparison.
2. Added compressor initialization logging showing the detected
context_length, threshold, model, provider, and base_url.
This makes turn-1 compression bugs diagnosable from logs —
previously there was no log of what context length was detected.
When using Alibaba (DashScope) with an anthropic-compatible endpoint,
model names like qwen3.5-plus were being normalized to qwen3-5-plus.
Alibaba's API expects the dot. Added preserve_dots parameter to
normalize_model_name() and build_anthropic_kwargs().
Also fixed 401 auth: when provider is alibaba or base_url contains
dashscope/aliyuncs, use only the resolved API key (DASHSCOPE_API_KEY).
Never fall back to resolve_anthropic_token(), and skip Anthropic
credential refresh for DashScope endpoints.
Cherry-picked from PR #1748 by crazywriter1. Fixes#1739.
Six improvements to reduce information loss during context compression,
informed by analysis of Cline, OpenCode, Pi-mono, Codex, and ClawdBot:
1. Structured summary template — sections for Goal, Progress (Done/
In Progress/Blocked), Key Decisions, Relevant Files, Next Steps,
and Critical Context. Forces the summarizer to preserve each
category instead of writing a vague paragraph.
2. Iterative summary updates — on re-compression, the prompt says
'PRESERVE existing info, ADD new progress, UPDATE done/in-progress
status.' Previous summary is stored and fed back to the summarizer
so accumulated context survives across multiple compactions.
3. Token-budget tail protection — instead of fixed protect_last_n=4,
walks backward keeping ~20K tokens of recent context. Adapts to
message density: sessions with big tool results protect fewer
messages, short exchanges protect more. Falls back to protect_last_n
for small conversations.
4. Tool output pruning (pre-pass) — before the expensive LLM summary,
replaces old tool result contents with a placeholder. This is free
(no LLM call) and can save 30%+ of context by itself.
5. Scaled summary budget — instead of fixed 2500 tokens, allocates 20%
of compressed content tokens (clamped to 2000-8000). A 50-turn
conversation gets more summary space than a 10-turn one.
6. Richer summarizer input — tool calls now include arguments (up to
500 chars) and tool results keep up to 3000 chars (was 1500).
The summarizer sees 'terminal(git status) → M src/config.py'
instead of just '[Tool calls: terminal]'.
Previously, all project context files (AGENTS.md, .cursorrules, .hermes.md)
were loaded and concatenated into the system prompt. This bloated the prompt
with potentially redundant or conflicting instructions.
Now only ONE project context type is loaded, using priority order:
1. .hermes.md / HERMES.md (walk to git root)
2. AGENTS.md / agents.md (recursive directory walk)
3. CLAUDE.md / claude.md (cwd only, NEW)
4. .cursorrules / .cursor/rules/*.mdc (cwd only)
SOUL.md from HERMES_HOME remains independent and always loads.
Also adds CLAUDE.md as a recognized context file format, matching the
convention popularized by Claude Code.
Refactored the monolithic function into four focused helpers:
_load_hermes_md, _load_agents_md, _load_claude_md, _load_cursorrules.
Tests: replaced 1 coexistence test with 10 new tests covering priority
ordering, CLAUDE.md loading, case sensitivity, injection blocking.
In Docker/systemd/piped environments, the KawaiiSpinner animation
generates ~500 log lines per tool call. Now checks isatty() and
falls back to clean [tool]/[done] log lines in non-TTY contexts.
Interactive CLI behavior unchanged.
Based on work by 42-evey in PR #2203.
The official international DashScope endpoint uses dashscope-intl.aliyuncs.com
(per Alibaba docs), which the substring match on dashscope.aliyuncs.com misses
because of the hyphenated prefix.
If a tool_calls list contains a None entry (from malformed API response,
compression artifact, or corrupt session replay), convert_messages_to_anthropic
crashes with AttributeError: 'NoneType' object has no attribute 'get'.
Skip None and non-dict entries in the tool_calls iteration. Found via
chaos/fuzz testing with mixed valid/invalid tool_call entries.
Custom endpoint users (DashScope/Alibaba, Z.AI, Kimi, DeepSeek, etc.)
get wrong context lengths because their provider resolves as "openrouter"
or "custom", skipping the models.dev lookup entirely. For example,
qwen3.5-plus on DashScope falls to the generic "qwen" hardcoded default
(131K) instead of the correct 1M.
Add _infer_provider_from_url() that maps known API hostnames to their
models.dev provider IDs. When the explicit provider is generic
(openrouter/custom/empty), infer from the base URL before the models.dev
lookup. This resolves context lengths correctly for DashScope, Z.AI,
Kimi, MiniMax, DeepSeek, and Nous endpoints without requiring users to
manually set context_length in config.
Also refactors _is_known_provider_base_url() to use the same URL mapping,
removing the duplicated hostname list.
Cherry-picked from PR #2146 by @crazywriter1. Fixes#2104.
asyncio.run() creates and closes a fresh event loop each call. Cached
httpx/AsyncOpenAI clients bound to the dead loop crash on GC with
'Event loop is closed'. This hit vision_analyze on first use in CLI.
Two-layer fix:
- model_tools._run_async(): replace asyncio.run() with persistent
loop via _get_tool_loop() + run_until_complete()
- auxiliary_client._get_cached_client(): track which loop created
each async client, discard stale entries if loop is closed
6 regression tests covering loop lifecycle, reuse, and full vision
dispatch chain.
Co-authored-by: Test <test@test.com>
Cherry-picked from PR #2169 by @0xbyt4.
1. _strip_provider_prefix: skip Ollama model:tag names (qwen:0.5b)
2. Fuzzy match: remove reverse direction that made claude-sonnet-4
resolve to 1M instead of 200K
3. _has_content_after_think_block: reuse _strip_think_blocks() to
handle all tag variants (thinking, reasoning, REASONING_SCRATCHPAD)
4. models.dev lookup: elif→if so nous provider also queries models.dev
5. Disk cache fallback: use 5-min TTL instead of full hour so network
is retried soon
6. Delegate build: wrap child construction in try/finally so
_last_resolved_tool_names is always restored on exception
Two fixes for Telegram/gateway-specific bugs:
1. Anthropic adapter: strip orphaned tool_result blocks (mirror of
existing tool_use stripping). Context compression or session
truncation can remove an assistant message containing a tool_use
while leaving the subsequent tool_result intact. Anthropic rejects
these with a 400: 'unexpected tool_use_id found in tool_result
blocks'. The adapter now collects all tool_use IDs and filters out
any tool_result blocks referencing IDs not in that set.
2. Gateway: /reset and /new now bypass the running-agent guard (like
/status already does). Previously, sending /reset while an agent
was running caused the raw text to be queued and later fed back as
a user message with the same broken history — replaying the
corrupted session instead of resetting it. Now the running agent is
interrupted, pending messages are cleared, and the reset command
dispatches immediately.
Tests updated: existing tests now include proper tool_use→tool_result
pairs; two new tests cover orphaned tool_result stripping.
Co-authored-by: Test <test@test.com>
* feat: context pressure warnings for CLI and gateway
User-facing notifications as context approaches the compaction threshold.
Warnings fire at 60% and 85% of the way to compaction — relative to
the configured compression threshold, not the raw context window.
CLI: Formatted line with a progress bar showing distance to compaction.
Cyan at 60% (approaching), bold yellow at 85% (imminent).
◐ context ▰▰▰▰▰▰▰▰▰▰▰▰▱▱▱▱▱▱▱▱ 60% to compaction 100k threshold (50%) · approaching compaction
⚠ context ▰▰▰▰▰▰▰▰▰▰▰▰▰▰▰▰▰▱▱▱ 85% to compaction 100k threshold (50%) · compaction imminent
Gateway: Plain-text notification sent to the user's chat via the new
status_callback mechanism (asyncio.run_coroutine_threadsafe bridge,
same pattern as step_callback).
Does NOT inject into the message stream. The LLM never sees these
warnings. Flags reset after each compaction cycle.
Files changed:
- agent/display.py — format_context_pressure(), format_context_pressure_gateway()
- run_agent.py — status_callback param, _context_50/70_warned flags,
_emit_context_pressure(), flag reset in _compress_context()
- gateway/run.py — _status_callback_sync bridge, wired to AIAgent
- tests/test_context_pressure.py — 23 tests
* Merge remote-tracking branch 'origin/main' into hermes/hermes-7ea545bf
---------
Co-authored-by: Test <test@test.com>
Replace the fragile hardcoded context length system with a multi-source
resolution chain that correctly identifies context windows per provider.
Key changes:
- New agent/models_dev.py: Fetches and caches the models.dev registry
(3800+ models across 100+ providers with per-provider context windows).
In-memory cache (1hr TTL) + disk cache for cold starts.
- Rewritten get_model_context_length() resolution chain:
0. Config override (model.context_length)
1. Custom providers per-model context_length
2. Persistent disk cache
3. Endpoint /models (local servers)
4. Anthropic /v1/models API (max_input_tokens, API-key only)
5. OpenRouter live API (existing, unchanged)
6. Nous suffix-match via OpenRouter (dot/dash normalization)
7. models.dev registry lookup (provider-aware)
8. Thin hardcoded defaults (broad family patterns)
9. 128K fallback (was 2M)
- Provider-aware context: same model now correctly resolves to different
context windows per provider (e.g. claude-opus-4.6: 1M on Anthropic,
128K on GitHub Copilot). Provider name flows through ContextCompressor.
- DEFAULT_CONTEXT_LENGTHS shrunk from 80+ entries to ~16 broad patterns.
models.dev replaces the per-model hardcoding.
- CONTEXT_PROBE_TIERS changed from [2M, 1M, 512K, 200K, 128K, 64K, 32K]
to [128K, 64K, 32K, 16K, 8K]. Unknown models no longer start at 2M.
- hermes model: prompts for context_length when configuring custom
endpoints. Supports shorthand (32k, 128K). Saved to custom_providers
per-model config.
- custom_providers schema extended with optional models dict for
per-model context_length (backward compatible).
- Nous Portal: suffix-matches bare IDs (claude-opus-4-6) against
OpenRouter's prefixed IDs (anthropic/claude-opus-4.6) with dot/dash
normalization. Handles all 15 current Nous models.
- Anthropic direct: queries /v1/models for max_input_tokens. Only works
with regular API keys (sk-ant-api*), not OAuth tokens. Falls through
to models.dev for OAuth users.
Tests: 5574 passed (18 new tests for models_dev + updated probe tiers)
Docs: Updated configuration.md context length section, AGENTS.md
Co-authored-by: Test <test@test.com>
Cron jobs run unattended with no user present. Previously the agent had
send_message and clarify tools available, which makes no sense — the
final response is auto-delivered, and there's nobody to ask questions to.
Changes:
- Disable messaging and clarify toolsets for cron agent sessions
- Update cron platform hint to emphasize autonomous execution: no user
present, cannot ask questions, must execute fully and make decisions
- Update cronjob tool schema description to match (remove stale
send_message guidance)
* fix: preserve Ollama model:tag colons in context length detection
The colon-split logic in get_model_context_length() and
_query_local_context_length() assumed any colon meant provider:model
format (e.g. "local:my-model"). But Ollama uses model:tag format
(e.g. "qwen3.5:27b"), so the split turned "qwen3.5:27b" into just
"27b" — which matches nothing, causing a fallback to the 2M token
probe tier.
Now only recognised provider prefixes (local, openrouter, anthropic,
etc.) are stripped. Ollama model:tag names pass through intact.
* fix: update claude-opus-4-6 and claude-sonnet-4-6 context length from 200K to 1M
Both models support 1,000,000 token context windows. The hardcoded defaults
were set before Anthropic expanded the context for the 4.6 generation.
Verified via models.dev and OpenRouter API data.
---------
Co-authored-by: kshitijk4poor <82637225+kshitijk4poor@users.noreply.github.com>
Co-authored-by: Test <test@test.com>
The colon-split logic in get_model_context_length() and
_query_local_context_length() assumed any colon meant provider:model
format (e.g. "local:my-model"). But Ollama uses model:tag format
(e.g. "qwen3.5:27b"), so the split turned "qwen3.5:27b" into just
"27b" — which matches nothing, causing a fallback to the 2M token
probe tier.
Now only recognised provider prefixes (local, openrouter, anthropic,
etc.) are stripped. Ollama model:tag names pass through intact.
Co-authored-by: kshitijk4poor <82637225+kshitijk4poor@users.noreply.github.com>
Custom endpoints (LM Studio, Ollama, vLLM, llama.cpp) silently fall
back to 2M tokens when /v1/models doesn't include context_length.
Adds _query_local_context_length() which queries server-specific APIs:
- LM Studio: /api/v1/models (max_context_length + loaded instances)
- Ollama: /api/show (model_info + num_ctx parameters)
- llama.cpp: /props (n_ctx from default_generation_settings)
- vLLM: /v1/models/{model} (max_model_len)
Prefers loaded instance context over max (e.g., 122K loaded vs 1M max).
Results are cached via save_context_length() to avoid repeated queries.
Also fixes detect_local_server_type() misidentifying LM Studio as
Ollama (LM Studio returns 200 for /api/tags with an error body).
When LM Studio has a model loaded with a custom context size (e.g.,
122K), prefer that over the model's max_context_length (e.g., 1M).
This makes the TUI status bar show the actual runtime context window.
Instead of defaulting to 2M for unknown local models, query the server
API for the real context length. Supports Ollama (/api/show), vLLM
(max_model_len), and LM Studio (/v1/models). Results are cached to
avoid repeated queries.
Closes#1911
- insights.py: Pre-compute SELECT queries as class constants instead of
f-string interpolation at runtime. _SESSION_COLS is now evaluated once
at class definition time.
- hermes_state.py: Add identifier quoting and whitelist validation for
ALTER TABLE column names in schema migrations.
- Add 4 tests verifying no injection vectors in SQL query construction.
* fix: detect context length for custom model endpoints via fuzzy matching + config override
Custom model endpoints (non-OpenRouter, non-known-provider) were silently
falling back to 2M tokens when the model name didn't exactly match what the
endpoint's /v1/models reported. This happened because:
1. Endpoint metadata lookup used exact match only — model name mismatches
(e.g. 'qwen3.5:9b' vs 'Qwen3.5-9B-Q4_K_M.gguf') caused a miss
2. Single-model servers (common for local inference) required exact name
match even though only one model was loaded
3. No user escape hatch to manually set context length
Changes:
- Add fuzzy matching for endpoint model metadata: single-model servers
use the only available model regardless of name; multi-model servers
try substring matching in both directions
- Add model.context_length config override (highest priority) so users
can explicitly set their model's context length in config.yaml
- Log an informative message when falling back to 2M probe, telling
users about the config override option
- Thread config_context_length through ContextCompressor and AIAgent init
Tests: 6 new tests covering fuzzy match, single-model fallback, config
override (including zero/None edge cases).
* fix: auto-detect local model name and context length for local servers
Cherry-picked from PR #2043 by sudoingX.
- Auto-detect model name from local server's /v1/models when only one
model is loaded (no manual model name config needed)
- Add n_ctx_train and n_ctx to context length detection keys for llama.cpp
- Query llama.cpp /props endpoint for actual allocated context (not just
training context from GGUF metadata)
- Strip .gguf suffix from display in banner and status bar
- _auto_detect_local_model() in runtime_provider.py for CLI init
Co-authored-by: sudo <sudoingx@users.noreply.github.com>
* fix: revert accidental summary_target_tokens change + add docs for context_length config
- Revert summary_target_tokens from 2500 back to 500 (accidental change
during patching)
- Add 'Context Length Detection' section to Custom & Self-Hosted docs
explaining model.context_length config override
---------
Co-authored-by: Test <test@test.com>
Co-authored-by: sudo <sudoingx@users.noreply.github.com>
After #1675 removed ANTHROPIC_BASE_URL env var support, the Anthropic
provider base URL was hardcoded to https://api.anthropic.com. Now reads
model.base_url from config.yaml as an override, falling back to the
default when not set. Also applies to the auxiliary client.
Cherry-picked from PR #1949 by @rivercrab26.
Co-authored-by: rivercrab26 <rivercrab26@users.noreply.github.com>
_align_boundary_backward only checked messages[idx-1] to decide if
the compress-end boundary splits a tool_call/result group. When an
assistant issues 3+ parallel tool calls, their results span multiple
consecutive messages. If the boundary fell in the middle of that group,
the parent assistant was summarized away and orphaned tool results were
silently deleted by _sanitize_tool_pairs.
Now walks backward through all consecutive tool results to find the
parent assistant, then pulls the boundary before the entire group.
6 regression tests added in tests/test_compression_boundary.py.
Co-authored-by: Guts <Gutslabs@users.noreply.github.com>
SOUL.md now loads in slot #1 of the system prompt, replacing the
hardcoded DEFAULT_AGENT_IDENTITY. This lets users fully customize
the agent's identity and personality by editing ~/.hermes/SOUL.md
without it conflicting with the built-in identity text.
When SOUL.md is loaded as identity, it's excluded from the context
files section to avoid appearing twice. When SOUL.md is missing,
empty, unreadable, or skip_context_files is set, the hardcoded
DEFAULT_AGENT_IDENTITY is used as a fallback.
The default SOUL.md (seeded on first run) already contains the full
Hermes personality, so existing installs are unaffected.
Co-authored-by: Test <test@test.com>
