feat(cognitive-assembly-line): implement complete pipeline with 16 new MCP subagents, explore/orchestrator agents, and pipeline commands#159
Open
Delqhi wants to merge 15 commits into
Conversation
added 15 commits
May 4, 2026 13:59
…w MCP subagents, explore/orchestrator agents, and pipeline commands - Add 16 new MCP servers for Code Quality, Docs, DevOps, Data Science subagents - Integrate 21 subagents from infra-opencode-stack (hermes, prometheus, zeus, atlas, etc.) - Add explore agent (Step 3.5 Flash, reasoning high) for codebase analysis - Add orchestrator agent for Cognitive Assembly Line coordination - Add 8 pipeline commands (descriptor, router, validation, execution, aggregation, full, dispatch) - Enable sin-research MCP for web research - Document benchmark comparisons (GLM-5.1, DeepSeek V4, Qwen 3.6, MiniMax, Mistral) - Create cognitive-assembly-line-subagent-blueprint.md with complete subagent spec - Test pipeline-descriptor command successfully This sets up the complete Cognitive Assembly Line infrastructure for automated parallel agent workflows. Refs: #1 (Benchmark research issue created and dispatched)
…e commands; reset default_agent to SIN-Zeus
…us, athena, argus, zeus, omoc) instead of non-existent 'explore', 'librarian', 'oracle'
…peline - Add 15 functional MCP server scripts (bash-based) in .opencode/mcp-servers/ - Enable all MCPs in .opencode/opencode.json (enabled: true) - Update commands to use repo-relative paths for portability - Fix bash syntax errors (function definitions) - MCPs: code-checker, test-runner, security-scanner, performance-auditor, doc-writer, pr-generator, changelog-writer, ci-agent, env-manager, infra-provisioner, backup-agent, data-viz, data-analyzer, ml-trainer, ml-deployer - SIN-Zeus prompt already corrected in previous commit to use actual subagents (daedalus, athena, argus, zeus, omoc) These MCPs provide the foundation for the validation and execution layers of the Cognitive Assembly Line pipeline.
- code-checker: Add mypy, black, tsc, prettier support - test-runner: pytest integration - security-scanner: semgrep SAST - performance-auditor: lighthouse (fallback if missing) - Proper JSON-RPC error handling and structured output
- PIPELINE-ROUTER-MANIFEST.json: 5-route pipeline manifest (R-001 to R-005) - ULTRA-PLAN_FACTORIAL_IMPLEMENTATION.json: full ultra-plan with research, risk assessment, quality gates, and OKRs - Baseline SHA for Hermes dispatch reference: 9a0102c Artifacts attributable to SIN-Zeus pipeline-execution dispatch.
MiniMax M2.7 appeared twice in the model selector because it was defined in both nvidia-nim and fireworks-ai providers. Removed the nvidia-nim entry since all agents are configured to use fireworks-ai/minimax-m2.7.
- reasoningEffort: 'xhigh' -> 'high' (invalid value for MiniMax M2.7, only accepts low/medium/high per Fireworks API reference) - MiniMax M2.7 context: 262144 -> 196608 (196.6k per fireworks.ai) - Kimi K2.6: add modalities (supports text+image input) - Qwen 3.6 Plus: add modalities (vision model) + reduce context from 262144 to 131072 (no official context limit listed by Fireworks)
…@ai-sdk/fireworks provider The @ai-sdk/fireworks provider (v2.0.52) natively supports: - thinking/reasoning via providerOptions.fireworks.thinking - reasoningHistory (disabled/interleaved/preserved) - Proper model capability detection for reasoning/thinking UI controls Previously using @ai-sdk/openai-compatible which doesn't declare reasoning capabilities, causing OpenCode to hide thinking options for models like Kimi K2.6 and MiniMax M2.7.
- kimi-k2p6 (was kimi-k2p6 with wrong ID format) - minimax-m2p7 (was minimax-m2p7 with wrong prefix) - qwen3p6-plus (was qwen3p6-plus with wrong prefix) - Added kimi-k2-thinking as bonus Model IDs corrected to match actual Fireworks AI model URLs: - fireworks/kimi-k2p6 (262144 context, vision) - fireworks/minimax-m2p7 (196608 context) - fireworks/qwen3p6-plus (vision) - fireworks/kimi-k2-thinking (262144 context) Removed invalid reasoningEffort options from agents. Also fixed agent model references from minimax-m2.7 to minimax-m2p7.
DeepSeek V4 Pro: fireworks/deepseek-ai/deepseek-v4-pro - 1048576 context (1M tokens!) - 65536 output - $1.74/M Input • $3.48/M Output
GLM 5.1: fireworks/glm-5p1 (743.9B MoE, 202.8k context, $1.40/M input)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR implements the complete Cognitive Assembly Line infrastructure:
Refs: #1 (Benchmark research issue created and dispatched)
Related: Infra-SIN-OpenCode-Stack PR #X (oh-my-opencode.json integration)