scriptoriumadmin/fabro
1
0
Fork 0
mirror of https://github.com/fabro-sh/fabro.git synced 2026-10-08 03:10:26 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 5
164 commits 1524 branches 178 tags 244 MiB
Commit graph

6 commits

Author SHA1 Message Date
Bryan Helmkamp
7ad750c5b6 Rename tool handler to script handler and add language attribute 
Rename the entire "tool" concept to "script": ToolHandler → ScriptHandler,
tool_command attribute → script, tool.output → script.output, and all
related artifact filenames. Add a language attribute (shell | python,
default shell) so script nodes can run Python via python3 -c natively.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-02-27 09:50:39 -05:00
Bryan Helmkamp
1e3726b7f3 Add missing data to agent events and cost tracking in pipeline logs 
Enrich three AgentEvent variants with payload data (UserInput.text,
TurnLimitReached.max_turns, SteeringInjected.text) and forward
UserInput/SteeringInjected to the pipeline event stream. Add cost
field to StageUsage (computed from model pricing catalog) and
total_cost to PipelineCompleted (summed from stage outcomes).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Entire-Checkpoint: adb2363c9ec4
 2026-02-25 17:17:13 -05:00
Bryan Helmkamp
74a6b219cc Fix spec: FAIL does not trigger retries, only RETRY does 
The test criteria incorrectly said FAIL outcomes should be retried.
The pseudocode (Section 3.5) and the C reference implementation both
return immediately on FAIL. Only RETRY status triggers the retry loop.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-02-23 16:19:10 -05:00
Bryan Helmkamp
30cf3a6787 Rename coding-agent-loop crate to agent 
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-02-23 09:17:48 -05:00
Bryan Helmkamp
a2f9e87b7c Close spec compliance gaps: retry context, thread resolution, fan-in logging 
Fix 3 confirmed gaps from spec compliance review (85 items, 91.8% aligned):

- Write internal.retry_count.<node_id> to PipelineContext after retries
  so handlers and conditions can access retry counts (spec 5.1)
- Add graph-level default_thread (step 3) to 5-step thread ID resolution,
  pass graph param to resolve_thread_id (spec 5.4)
- Write prompt.md/response.md in fan_in LLM evaluation path (spec 5.6)

Also includes pre-existing improvements: checkpoint stores node_outcomes
and next_node_id for correct resume, engine timeout enforcement,
auto_status support, fidelity degradation on resume, preamble injection,
is_retryable error classification, stylesheet specificity correction,
full stylesheet parse validation, direction_valid lint rule, pre-hook
returns Skipped not Fail, fan-in score-based sorting and all-fail
detection, manager_loop child autostart and steer cooldown.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-02-22 10:49:21 -04:00
Bryan Helmkamp
40be740949 specs 2026-02-19 13:41:55 -04:00