AI & Machine Learning

AI & Machine Learning

AI predictions, model training, reinforcement learning, and RAG knowledge base.

12 tables in this group.

predictions

AI model predictions for validation and live monitoring

RLS Policies:

• Service role full access predictions — ALL for {public}
• Users can insert own predictions — INSERT for {public}
• Users can view own predictions — SELECT for {public}

Indexes:

• idx_predictions_agent_id
• idx_predictions_confidence
• idx_predictions_created
• idx_predictions_model
• idx_predictions_outcome
• idx_predictions_pending_is_correct
• idx_predictions_pending_outcome
• idx_predictions_symbol
• idx_predictions_type
• idx_predictions_user
• idx_predictions_user_id
• predictions_pkey

---

model_training_jobs

ML model training runs (LSTM, transformers, RL agents)

RLS Policies:

• Service role full access model_training_jobs — ALL for {public}
• Users can insert own training jobs — INSERT for {public}
• Users can update own training jobs — UPDATE for {public}
• Users can view own training jobs — SELECT for {public}

Indexes:

• idx_model_training_jobs_created
• idx_model_training_jobs_model_name
• idx_model_training_jobs_model_type
• idx_model_training_jobs_status
• idx_model_training_jobs_user
• model_training_jobs_pkey

---

rl_training_episodes

Reinforcement learning episode data (Q-Learning, DT, PPO)

RLS Policies:

• Service role full access rl_training_episodes — ALL for {public}

Indexes:

• idx_rl_episodes_agent
• idx_rl_episodes_algorithm
• idx_rl_episodes_created
• idx_rl_episodes_number
• idx_rl_episodes_reward
• idx_rl_episodes_symbol
• idx_rl_training_episodes_agent_id
• rl_training_episodes_pkey

---

agent_learning_episodes

Agent self-improvement learning episodes

RLS Policies:

• Service role full access agent_learning_episodes — ALL for {public}

Indexes:

• agent_learning_episodes_pkey
• idx_agent_learning_agent
• idx_agent_learning_created
• idx_agent_learning_episodes_agent_id
• idx_agent_learning_outcome
• idx_agent_learning_reward
• idx_agent_learning_type

---

collective_memory

Universal shared memory for cross-agent knowledge transfer and collective learning

RLS Policies:

• All authenticated users can read collective memory — SELECT for {public}
• Service role full access to collective memory — ALL for {public}

Indexes:

• collective_memory_pkey
• idx_collective_memory_agent
• idx_collective_memory_agent_id
• idx_collective_memory_agent_symbol_type
• idx_collective_memory_confidence
• idx_collective_memory_created
• idx_collective_memory_embedding
• idx_collective_memory_expires
• idx_collective_memory_memory_type
• idx_collective_memory_symbol
• idx_collective_memory_tags
• idx_collective_memory_type

---

q_table_snapshots

Persisted Q-value snapshots from RL agents for cross-agent aggregation

RLS Policies:

• All authenticated users can read Q-table snapshots — SELECT for {public}
• Service role full access to Q-table snapshots — ALL for {public}

Indexes:

• idx_q_snapshots_agent
• idx_q_snapshots_created
• idx_q_snapshots_state
• idx_q_snapshots_symbol
• idx_q_snapshots_symbol_state_action
• idx_q_table_snapshots_upsert
• q_table_snapshots_pkey

---

trading_experiences

RL experiences for Decision Transformer and Q-Learning training

RLS Policies:

• Authenticated users can view experience statistics — SELECT for {public}
• Service role can manage all experiences — ALL for {public}

Indexes:

• trading_experiences_action_idx
• trading_experiences_agent_id_idx
• trading_experiences_agent_reward_idx
• trading_experiences_created_at_idx
• trading_experiences_embedding_idx
• trading_experiences_pkey
• trading_experiences_reward_idx

---

trading_documents

RAG knowledge base for trading content with semantic embeddings for similarity search

RLS Policies:

• trading_documents_delete_policy — DELETE for {authenticated}
• trading_documents_insert_policy — INSERT for {authenticated}
• trading_documents_read_policy — SELECT for {authenticated}
• trading_documents_service_role_policy — ALL for {service_role}
• trading_documents_update_policy — UPDATE for {authenticated}

Indexes:

• trading_documents_content_created_at_idx
• trading_documents_content_hash_key
• trading_documents_created_at_idx
• trading_documents_doc_type_idx
• trading_documents_embedding_hnsw_idx
• trading_documents_expires_at_idx
• trading_documents_metadata_gin_idx
• trading_documents_pkey
• trading_documents_source_idx
• trading_documents_user_id_idx

---

rag_settings

Admin-configurable settings for RAG data collection pipeline. Should contain only one row (singleton pattern).

RLS Policies:

• Admins can create RAG settings — INSERT for {authenticated}
• Admins can delete RAG settings — DELETE for {authenticated}
• Admins can update RAG settings — UPDATE for {authenticated}
• Authenticated users can view RAG settings — SELECT for {authenticated}

Indexes:

• rag_settings_pkey
• rag_settings_updated_at_idx

---

ai_batch_jobs

RLS Policies:

• Service role full access to batch jobs — ALL for {public}
• Users can cancel own pending batch jobs — UPDATE for {public}
• Users can insert own batch jobs — INSERT for {public}
• Users can view own batch jobs — SELECT for {public}

Indexes:

• ai_batch_jobs_custom_id_key
• ai_batch_jobs_idempotency_key_key
• ai_batch_jobs_pkey
• idx_ai_batch_jobs_batch_id
• idx_ai_batch_jobs_job_type
• idx_ai_batch_jobs_pending
• idx_ai_batch_jobs_status
• idx_ai_batch_jobs_user_id

---

ai_batch_executions

RLS Policies:

• Service role full access to batch executions — ALL for {public}

Indexes:

• ai_batch_executions_anthropic_batch_id_key
• ai_batch_executions_pkey
• idx_ai_batch_executions_status

---

attribution_analyses

RLS Policies:

• attribution_analyses_service_all — ALL for {public}

Indexes:

• attribution_analyses_pkey
• idx_attribution_analyses_benchmark_id
• idx_attribution_analyses_portfolio_id

---