ID Title Status Priority Created Due Date Actions
#772 Live activity worker emits completion event without tenant_id (durable-leaf child runs hang for non-default tenants)
hwe activity-worker uses enterprise/services/activity_worker.py. Its _complete_activity/_fail_activi...
closed high 2026-06-29 00:09 -
#771 Agentic durable leaves crash on activity worker: ActivityContext has no jumper_client
durable_leaves=true routes tools.agent.execute_leaf_agent to the activity worker (agent_leaf DSL, 60...
closed critical 2026-06-29 00:04 -
#770 Agentic Theme A1: MCP client (tools.mcp.call) to call external MCP servers as durable tools
Roadmap Theme A1 (breadth unlock). New engine tools tools.mcp.call + tools.mcp.list_tools (engine/to...
open high 2026-06-22 16:11 -
Edit
#769 Codify engine/apps boundary + relocate use-case DSL templates out of served gallery
Holistic review (3 parallel reviewers: engine, api, enterprise) found the architecture clean: engine...
closed medium 2026-06-22 13:35 -
#768 Agent robustness: switch model to minimax-m3, add token-budget + gmail body ceiling, fail-fast reaper
Model A/B (gemma4:31b vs glm-5.2 vs minimax-m3) on an unanswerable question: gemma over-reached, bal...
closed high 2026-06-22 04:01 -
#767 Fix rate-limiter concurrency bug: savepoint-isolate tool check, fail-open tenant limiter, scale workers
ROOT CAUSE: the per-tool rate-limit decorator ran check_and_consume on the agent task own DB connect...
closed high 2026-06-22 03:01 -
#766 Comprehensive review fixes: webhook crash-safety, dedup, disconnect cleanup, security hardening
3-reviewer pass (security/correctness/consistency) fixes. CORRECTNESS: dispatch_reaction never raise...
closed high 2026-06-22 01:59 -
#765 Telegram connector + real-time triggers on the reusable reaction framework
Second channel, validated e2e. Net-new only: TelegramApp (enterprise/apps/pre_installed/telegram.py,...
closed high 2026-06-22 01:12 -
#764 Phase B-2: Gmail real-time triggers via Cloud Pub/Sub — ingestion validated
Gmail users.watch -> Pub/Sub topic (projects/highway-481210/topics/Gmail+Highway) -> push to POST /a...
open high 2026-06-22 00:06 -
Edit
#763 Phase B: Gmail triggers (watch_gmail) — durable poll-and-react
watch_gmail(match, instruction, interval_minutes) + stop_gmail_watch(watch_id) MCP tools. Reuses the...
closed high 2026-06-21 23:27 -
#762 Gmail connector (Highway app) + managed per-tenant OAuth + Vault-backed config
First real connector, end-to-end on demo. (1) ENGINE FIX: docker/highway-core/engine/config.py get_s...
closed high 2026-06-21 22:45 -
#761 Engine: workflow result empty after sleep/resume (deferred get_status shows {})
_select_workflow_result (inline_executor/core.py:210) returns the LAST task's {task}.result variable...
closed medium 2026-06-21 03:38 -
#760 Security: network-isolated sandbox for tools.shell.run (contained host/DB/secrets)
Wired tools.shell.run to the enterprise Docker sandbox with network_mode=none. run_shell_in_sandbox ...
closed high 2026-06-21 03:07 -
#759 Deploy: Highway MCP server public at https://mcp.highway.rodmena.app (multi-tenant)
Published the Highway MCP server as a public, multi-tenant HTTPS endpoint. (1) DNS: mcp.highway.rodm...
closed high 2026-06-21 02:23 -
#758 Scheduling: MCP schedule_goal + durable_cron for recurring/scheduled agentic workflows
Added recurring scheduling via Highway durable_cron. New no-HITL scheduled-agent definition (agent_s...
closed high 2026-06-21 01:32 -
#757 Agent items 2+3: idempotency (#750 partial), prompt-injection defense, cost/budget metering, eval harness
run_goal hardening (validated on demo): (1) COST METERING: token_budget param + per-run token accoun...
closed high 2026-06-21 01:06 -
#756 Agent tool catalog: wire http/db/email (+shell) into run_goal with framework-injected credentials
Added enterprise/tools/agent_catalog.py build_agent_tools(catalog, config): maps friendly keys (shel...
closed high 2026-06-21 00:52 -
#755 CRITICAL: HITL approval bypassed for 2nd+ approval in one task (request_approval wait checkpoint keyed by task id)
enterprise/tools/approval_tool.py request_approval used step_name=_current_task_id (the DSL task id)...
closed critical 2026-06-21 00:30 -
#754 Drop functiongemma:270m as agent function model; use gemma4:31b-cloud (270M too weak for multi-step)
Empirical (demo tenant): functiongemma:270m emits malformed/empty tool args on compound prompts (run...
open high 2026-06-21 00:17 -
Edit
#753 run_goal: empty final answer (gemma returns no text after tool-call history replay)
Validated Phase 0 dogfood (runs d00b9072 auto, 8356369b HITL) complete successfully but agent_result...
closed medium 2026-06-20 23:40 -
Previous Page 8 of 45 Next