common/hermes
2
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

feat(api): structured run events via /v1/runs SSE endpoint

Browse Source 
Add POST /v1/runs to start async agent runs and GET /v1/runs/{run_id}/events
for SSE streaming of typed lifecycle events (tool.started, tool.completed,
message.delta, reasoning.available, run.completed, run.failed).

Changes the internal tool_progress_callback signature from positional
(tool_name, preview, args) to event-type-first
(event_type, tool_name, preview, args, **kwargs). Existing consumers
filter on event_type and remain backward-compatible.

Adds concurrency limit (_MAX_CONCURRENT_RUNS=10) and orphaned run sweep.

Fixes logic inversion in cli.py _on_tool_progress where the original PR
would have displayed internal tools instead of non-internal ones.

Co-authored-by: Mibayy <mibayy@users.noreply.github.com>
This commit is contained in:
Mibayy
2026-04-05 11:52:46 -07:00
committed by Teknium
parent e167ad8f61
commit cc2b56b26a
11 changed files with 337 additions and 44 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
run_agent.py
View File

@@ -6056,7 +6056,7 @@ class AIAgent:
if self.tool_progress_callback:
try:
preview = _build_tool_preview(name, args)
self.tool_progress_callback(name, preview, args)
self.tool_progress_callback("tool.started", name, preview, args)
except Exception as cb_err:
logging.debug(f"Tool progress callback error: {cb_err}")
@@ -6121,6 +6121,15 @@ class AIAgent:
result_preview = function_result[:200] if len(function_result) > 200 else function_result
logger.warning("Tool %s returned error (%.2fs): %s", function_name, tool_duration, result_preview)
if self.tool_progress_callback:
try:
self.tool_progress_callback(
"tool.completed", function_name, None, None,
duration=tool_duration, is_error=is_error,
)
except Exception as cb_err:
logging.debug(f"Tool progress callback error: {cb_err}")
if self.verbose_logging:
logging.debug(f"Tool {function_name} completed in {tool_duration:.2f}s")
logging.debug(f"Tool result ({len(function_result)} chars): {function_result}")
@@ -6220,7 +6229,7 @@ class AIAgent:
if self.tool_progress_callback:
try:
preview = _build_tool_preview(function_name, function_args)
self.tool_progress_callback(function_name, preview, function_args)
self.tool_progress_callback("tool.started", function_name, preview, function_args)
except Exception as cb_err:
logging.debug(f"Tool progress callback error: {cb_err}")
@@ -6407,6 +6416,15 @@ class AIAgent:
if _is_error_result:
logger.warning("Tool %s returned error (%.2fs): %s", function_name, tool_duration, result_preview)
if self.tool_progress_callback:
try:
self.tool_progress_callback(
"tool.completed", function_name, None, None,
duration=tool_duration, is_error=_is_error_result,
)
except Exception as cb_err:
logging.debug(f"Tool progress callback error: {cb_err}")
if self.verbose_logging:
logging.debug(f"Tool {function_name} completed in {tool_duration:.2f}s")
logging.debug(f"Tool result ({len(function_result)} chars): {function_result}")
@@ -8263,21 +8281,25 @@ class AIAgent:
# Notify progress callback of model's thinking (used by subagent
# delegation to relay the child's reasoning to the parent display).
# Guard: only fire for subagents (_delegate_depth >= 1) to avoid
# spamming gateway platforms with the main agent's every thought.
if (assistant_message.content and self.tool_progress_callback
and getattr(self, '_delegate_depth', 0) > 0):
if (assistant_message.content and self.tool_progress_callback):
_think_text = assistant_message.content.strip()
# Strip reasoning XML tags that shouldn't leak to parent display
_think_text = re.sub(
r'</?(?:REASONING_SCRATCHPAD|think|reasoning)>', '', _think_text
).strip()
# For subagents: relay first line to parent display (existing behaviour).
# For all agents with a structured callback: emit reasoning.available event.
first_line = _think_text.split('\n')[0][:80] if _think_text else ""
if first_line:
if first_line and getattr(self, '_delegate_depth', 0) > 0:
try:
self.tool_progress_callback("_thinking", first_line)
except Exception:
pass
elif _think_text:
try:
self.tool_progress_callback("reasoning.available", "_thinking", _think_text[:500], None)
except Exception:
pass
# Check for incomplete <REASONING_SCRATCHPAD> (opened but never closed)
# This means the model ran out of output tokens mid-reasoning — retry up to 2 times