dexorder/ai
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

feat: add @tag model override support and remove Qdrant dependencies

Browse Source 
- Add model-tags parser for @Tag syntax in chat messages
- Support Anthropic models (Sonnet, Haiku, Opus) via @tag
- Remove Qdrant vector database from infrastructure and configs
- Simplify license model config to use null fallbacks
- Add greeting stream after model switch via @tag
- Fix protobuf field names to camelCase for v7 compatibility
- Add 429 rate limit retry logic with exponential backoff
- Remove RAG references from agent harness documentation
This commit is contained in:
Tim Olson
2026-04-27 20:55:18 -04:00
parent 6f937f9e5e
commit d41fcd0499
50 changed files with 956 additions and 798 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
doc/CLUSTER_SETUP.md
View File

@@ -10,7 +10,7 @@ The platform runs across two namespaces:
| Namespace | Contents |
|-----------|----------|
| `ai` | Gateway, web UI, all infrastructure services (postgres, minio, kafka, flink, relay, ingestor, qdrant, dragonfly, iceberg-catalog) |
| `ai` | Gateway, web UI, all infrastructure services (postgres, minio, kafka, flink, relay, ingestor, dragonfly, iceberg-catalog) |
| `sandbox` | Per-user sandbox containers (created dynamically by the gateway) |
Secrets are managed via 1Password CLI (`op inject`). All `.tpl.yaml` files in `deploy/k8s/prod/secrets/` contain `op://` references and are safe to commit; actual values are never stored in git.
@@ -217,7 +217,7 @@ kubectl --context=prod -n ai get configmaps
## Step 7 — Deploy Infrastructure
Infrastructure services (postgres, minio, kafka, iceberg-catalog, dragonfly, qdrant, relay, ingestor, flink) are defined in `deploy/k8s/prod/infrastructure.yaml` and were applied in Step 4.
Infrastructure services (postgres, minio, kafka, iceberg-catalog, dragonfly, relay, ingestor, flink) are defined in `deploy/k8s/prod/infrastructure.yaml` and were applied in Step 4.
Wait for the StatefulSets and Deployments to become ready:
@@ -225,7 +225,6 @@ Wait for the StatefulSets and Deployments to become ready:
kubectl --context=prod -n ai rollout status statefulset/postgres
kubectl --context=prod -n ai rollout status statefulset/minio
kubectl --context=prod -n ai rollout status statefulset/kafka
kubectl --context=prod -n ai rollout status statefulset/qdrant
kubectl --context=prod -n ai rollout status deployment/dragonfly
kubectl --context=prod -n ai rollout status deployment/iceberg-catalog
kubectl --context=prod -n ai rollout status deployment/relay

107
doc/agent_harness.md
View File

@@ -22,20 +22,20 @@ The Agent Harness is the core orchestration layer for the Dexorder AI platform,
│ ┌──────────────────┼──────────────────┐ │
│ │ │ │ │
│ ┌────▼─────┐ ┌────▼─────┐ ┌────▼─────┐ │
│ │ MCP │ │ LLM │ │ RAG │ │
│ │ Connector│ │ Router │ │ Retriever│ │
│ └────┬─────┘ └────┬─────┘ └────┬─────┘ │
│ │ │ │ │
└─────────┼──────────────────┼──────────────────┼─────────────┘
│ │ │
▼ ▼ ▼
┌────────────┐ ┌───────────┐ ┌───────────┐
│ User's │ │ LLM │ │ Qdrant │
│ MCP │ │ Providers │ │ (Vectors) │
│ Container │ │(Anthropic,│ │ │
│ (k8s pod) │ │ OpenAI, │ │ Global + │
│ │ │ etc) │ │ User │
└────────────┘ └───────────┘ └───────────┘
│ │ MCP │ │ LLM │ │
│ │ Connector│ │ Router │ │
│ └────┬─────┘ └────┬─────┘ │
│ │ │ │
└─────────┼──────────────────┼─────────────┘
│ │
▼ ▼
┌────────────┐ ┌───────────┐
│ User's │ │ LLM │
│ MCP │ │ Providers │
│ Container │ │(Anthropic,│
│ (k8s pod) │ │ OpenAI, │
│ │ │ etc) │
└────────────┘ └───────────┘
```
## Message Processing Flow
@@ -57,17 +57,11 @@ When a user sends a message:
│ - context://workspace-state
│ - context://system-prompt
│
├─→ b. RAGRetriever searches for relevant memories:
│ - Embeds user query
│ - Searches Qdrant: user_id = current_user OR user_id = "0"
│ - Returns user-specific + global platform knowledge
│
├─→ c. Build system prompt:
├─→ b. Build system prompt:
│ - Base platform prompt
│ - User profile context
│ - Workspace state
│ - Custom user instructions
│ - Relevant RAG memories
│
├─→ d. ModelRouter selects LLM:
│ - Based on license tier
@@ -92,11 +86,10 @@ When a user sends a message:
### 1. Agent Harness (`gateway/src/harness/agent-harness.ts`)
**Stateless orchestrator** - all state lives in user's MCP container or RAG.
**Stateless orchestrator** - all state lives in user's MCP container.
**Responsibilities:**
- Fetch context from user's MCP resources
- Query RAG for relevant memories
- Build prompts with full context
- Route to appropriate LLM
- Handle tool calls (platform vs user)
@@ -141,40 +134,12 @@ Routes queries to appropriate LLM based on:
- LangGraph checkpoints (1 hour TTL)
- Fast reads for active conversations
**Qdrant** (Vector Search)
- Conversation embeddings
- User-specific memories (user_id = actual user ID)
- **Global platform knowledge** (user_id = "0")
- RAG retrieval with cosine similarity
- GDPR-compliant (indexed by user_id for fast deletion)
**Iceberg** (Cold Storage)
- Full conversation history (partitioned by user_id, session_id)
- Checkpoint snapshots for replay
- Analytics and time-travel queries
- GDPR-compliant with compaction
#### RAG System:
**Global Knowledge** (user_id="0"):
- Platform capabilities and architecture
- Trading concepts and fundamentals
- Indicator development guides
- Strategy patterns and examples
- Loaded from `gateway/knowledge/` markdown files
**User Knowledge** (user_id=specific user):
- Personal conversation history
- Trading preferences and style
- Custom indicators and strategies
- Workspace state and context
**Query Flow:**
1. User query is embedded using EmbeddingService
2. Qdrant searches: `user_id IN (current_user, "0")`
3. Top-K relevant chunks returned
4. Added to LLM context automatically
### 5. Skills vs Subagents
#### Skills (`gateway/src/harness/skills/`)
@@ -290,44 +255,6 @@ User's MCP container provides access to:
- Tactical order generators (TWAP, iceberg, etc.)
- Smart order routing
## Global Knowledge Management
### Document Loading
At gateway startup:
1. DocumentLoader scans `gateway/knowledge/` directory
2. Markdown files chunked by headers (~1000 tokens/chunk)
3. Embeddings generated via EmbeddingService
4. Stored in Qdrant with user_id="0"
5. Content hashing enables incremental updates
### Directory Structure
```
gateway/knowledge/
├── platform/ # Platform capabilities
├── trading/ # Trading fundamentals
├── indicators/ # Indicator development
└── strategies/ # Strategy patterns
```
### Updating Knowledge
**Development:**
```bash
curl -X POST http://localhost:3000/admin/reload-knowledge
```
**Production:**
- Update markdown files
- Deploy new version
- Auto-loaded on startup
**Monitoring:**
```bash
curl http://localhost:3000/admin/knowledge-stats
```
## Container Lifecycle
### User Container Creation
@@ -362,7 +289,6 @@ When user connects:
### ✅ Completed
- Agent Harness with MCP integration
- Model routing with license tiers
- RAG retriever with Qdrant
- Document loader for global knowledge
- EmbeddingService (Ollama/OpenAI)
- Skills and subagents framework
@@ -388,5 +314,4 @@ When user connects:
- Documentation: `gateway/src/harness/README.md`
- Knowledge base: `gateway/knowledge/`
- LangGraph: https://langchain-ai.github.io/langgraphjs/
- Qdrant: https://qdrant.tech/documentation/
- MCP Spec: https://modelcontextprotocol.io/

25
doc/architecture.md
View File

@@ -19,7 +19,6 @@ Dexorder is an AI-powered trading platform that combines real-time market data p
│ • Authentication & session management │
│ • Agent Harness (LangChain/LangGraph orchestration) │
│ - MCP client connector to user containers │
│ - RAG retriever (Qdrant) │
│ - Model router (LLM selection) │
│ - Skills & subagents framework │
│ • Dynamic user container provisioning │
@@ -30,8 +29,7 @@ Dexorder is an AI-powered trading platform that combines real-time market data p
┌──────────────────┐ ┌──────────────┐ ┌──────────────────────┐
│ User Containers │ │ Relay │ │ Infrastructure │
│ (per-user pods) │ │ (ZMQ Router) │ │ • DragonflyDB (cache)│
│ │ │ │ │ • Qdrant (vectors) │
│ • MCP Server │ │ • Market data│ │ • PostgreSQL (meta) │
│ │ │ │ • MCP Server │ │ • Market data│ │ • PostgreSQL (meta) │
│ • User files: │ │ fanout │ │ • MinIO (S3) │
│ - Indicators │ │ • Work queue │ │ │
│ - Strategies │ │ • Stateless │ │ │
@@ -86,18 +84,16 @@ Dexorder is an AI-powered trading platform that combines real-time market data p
- **Agent Harness (LangChain/LangGraph):** ([[agent_harness]])
- Stateless LLM orchestration
- MCP client connector to user containers
- RAG retrieval from Qdrant (global + user-specific knowledge)
- Model routing based on license tier and complexity
- Skills and subagents framework
- Workflow state machines with validation loops
**Key Features:**
- **Stateless design:** All conversation state lives in user containers or Qdrant
- **Stateless design:** All conversation state lives in user containers
- **Multi-channel support:** WebSocket, Telegram (future: mobile, Discord, Slack)
- **Kubernetes-native:** Uses k8s API for container management
- **Three-tier memory:**
- Redis: Hot storage, active sessions, LangGraph checkpoints (1 hour TTL)
- Qdrant: Vector search, RAG, global + user knowledge, GDPR-compliant
- Iceberg: Cold storage, full history, analytics, time-travel queries
**Infrastructure:**
@@ -270,12 +266,6 @@ Exchange API → Ingestor → Kafka → Flink → Iceberg
- Redis-compatible in-memory cache
- Session state, rate limiting, hot data
#### Qdrant
- Vector database for RAG
- **Global knowledge** (user_id="0"): Platform capabilities, trading concepts, strategy patterns
- **User knowledge** (user_id=specific): Personal conversations, preferences, strategies
- GDPR-compliant (indexed by user_id for fast deletion)
#### PostgreSQL
- Iceberg catalog metadata
- User accounts and license info (gateway)
@@ -458,17 +448,11 @@ The gateway's agent harness (LangChain/LangGraph) orchestrates LLM interactions
│ - context://workspace-state
│ - context://system-prompt
│
├─→ b. RAGRetriever searches Qdrant for relevant memories:
│ - Embeds user query
│ - Searches: user_id IN (current_user, "0")
│ - Returns user-specific + global platform knowledge
│
├─→ c. Build system prompt:
├─→ b. Build system prompt:
│ - Base platform prompt
│ - User profile context
│ - Workspace state
│ - Custom user instructions
│ - Relevant RAG memories
│
├─→ d. ModelRouter selects LLM:
│ - Based on license tier
@@ -492,8 +476,6 @@ The gateway's agent harness (LangChain/LangGraph) orchestrates LLM interactions
**Key Architecture:**
- **Gateway is stateless:** No conversation history stored in gateway
- **User context in MCP:** All user-specific data lives in user's container
- **Global knowledge in Qdrant:** Platform documentation loaded from `gateway/knowledge/`
- **RAG at gateway level:** Semantic search combines global + user knowledge
- **Skills vs Subagents:**
- Skills: Well-defined, single-purpose tasks
- Subagents: Complex domain expertise with multi-file context
@@ -630,7 +612,6 @@ See [[backend_redesign]] for detailed notes.
- Historical backfill service
**Phase 3: Agent Features**
- RAG integration (Qdrant)
- Strategy backtesting
- Risk management tools
- Portfolio analytics

10
doc/plan.md
View File

@@ -14,3 +14,13 @@
* TradingView indicator import tool
* Results persistence: ~~research analysis~~, backtests, strategy performance metrics, etc.
* Free tier with token limits and sandbox shutdown
* Performance analysis
* Custom pre-session scanners / summaries
* Saved prompts (Create /presession prompt command for easy re-use)
https://github.com/wangzhe3224/awesome-systematic-trading
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3247865 151 trading strategies
https://vectorbt.dev/
https://github.com/shiyu-coder/Kronos
https://x.com/RohOnChain/status/2041180375838498950?s=20 combining signals