Deprecations and Migrations

Use this page to migrate old integrations to the current API contract.

Prerequisites

• Inventory of current payloads used by your app
• Access to SDK wrapper or API client code paths
• API reference available for schema checks

Minimal Setup

1. Replace deprecated fields with current equivalents.
2. Re-run end-to-end tests with baseline retrieval settings.
3. Update templates and SDK wrappers to emit only current fields.

Field Migrations

Search Ranking (policy → reranking_config) 🚨 BREAKING CHANGE

Effective: API v1.9.0+ (June 2026)

The search endpoint now uses reranking_config instead of policy.vector for controlling result ranking.

Deprecated Field on Search:

Reranking Providers:

• none - Cosine-only (no reranking)
• cohere - Cohere rerank-v3.5 (default)
• openai - OpenAI gpt-5-nano or gpt-5-mini
• papr_enhanced - Graph rerank (replaces old mode: "enhanced")
• papr_max - Graph rerank + CE + EGR (replaces old mode: "max")

Migration Examples:

Before (Deprecated):

# Search with policy.vector - OLD WAY
results = client.memory.search(
    query="How to sort a list?",
    policy={
        "vector": {
            "mode": "enhanced",  # DEPRECATED
            "domain_id": "cosqa",
            "signal_thresholds": {
                "language": 0.9
            }
        }
    }
)

After (Current):

# Search with reranking_config - NEW WAY
results = client.memory.search(
    query="How to sort a list?",
    reranking_config={
        "reranking_provider": "papr_enhanced",  # or "papr_max"
        "domain_id": "cosqa",
        "signal_thresholds": {
            "language": 0.9
        }
    }
)

Mode Mapping:

Graph-Aware Embeddings (Holographic → Graph API) 🚨 BREAKING CHANGE

Effective: API v1.8.0+ (May 2026)

All holographic-related fields and endpoints have been deprecated and replaced with a unified policy structure using graph transform APIs.

Deprecated Fields on Add/Update/Document:

Deprecated Field on Search:

Deprecated Endpoints:

Migration Examples:

Before (Deprecated):

# Add memory - OLD WAY
memory = client.memory.add(
    content="Python list sorting example",
    enable_holographic=True,
    frequency_schema_id="cosqa"
)

# Search - OLD WAY
results = client.memory.search(
    query="How to sort a list?",
    holographic_config={
        "enabled": True,
        "frequency_schema_id": "cosqa",
        "frequency_filters": {
            "language": 0.9
        }
    }
)

After (Current):

# Add memory - NEW WAY
memory = client.memory.add(
    content="Python list sorting example",
    policy={
        "transform_embedding": {
            "mode": "auto",
            "domain_id": "cosqa"
        }
    }
)

# Search - NEW WAY (v1.9.0+)
results = client.memory.search(
    query="How to sort a list?",
    reranking_config={
        "reranking_provider": "papr_enhanced",  # or "papr_max"
        "domain_id": "cosqa",
        "signal_thresholds": {
            "language": 0.9
        }
    }
)

Domain Registration Migration:

Before (Deprecated):

POST /v1/holographic/domains

{
  "name": "acme:legal:1.0.0",
  "description": "Legal contracts",
  "fields": [...]
}

After (Current):

POST /v1/graph/domains

{
  "domain_id": "acme:legal:1.0.0",
  "name": "Legal Contracts",
  "description": "Legal contract analysis",
  "signals": [...]
}

Key changes:

• name → domain_id (the unique identifier)
• name is now a human-readable display name
• fields → signals

Search Limit Increase:

The maximum number of results in /v1/memory/search has been increased from 50 to 200. Update any hardcoded limits or validation logic.

# OLD: max 50 results
results = client.memory.search(query="...", limit=50)

# NEW: max 200 results
results = client.memory.search(query="...", limit=200)

Graph Configuration

• Deprecated: graph_generation
• Current: memory_policy

Use memory_policy for:

• graph mode (auto, manual)
• extraction constraints
• governance (acl, consent, risk)

User Identity

• Deprecated: user_id (internal)
• Current: external_user_id (application identity)

Recommended: pass your app-level ID as external_user_id.

Search Ranking Toggle

• Deprecated: rank_results
• Current: reranking_config where applicable

For most integrations, prioritize:

• better query quality
• enable_agentic_graph=true
• tuned max_memories / max_nodes

Legacy Relationship Fields

• Deprecated or legacy patterns: relationships_json-first workflows
• Current: model graph behavior via memory_policy and link_to where possible

Migration Checklist

For Search Ranking (policy → reranking_config): ⚡ Latest

1. Update Search calls: Replace policy.vector with reranking_config
2. Update provider names: mode: "enhanced" → reranking_provider: "papr_enhanced"
3. Update mode mapping: See table above for complete mapping
4. Remove deprecated fields: Remove rank_results and holographic_config usage
5. Test reranking: Verify results with new provider names

For Graph-Aware Embeddings (Holographic → Graph):

1. Update Add/Update calls: Replace enable_holographic + frequency_schema_id with policy.transform_embedding
2. Update Search calls: Replace holographic_config with policy.vector Use reranking_config (see above)
3. Update domain registration: Use POST /v1/graph/domains instead of /v1/holographic/domains
4. Update field names: Change frequency_schema_id to domain_id, fields to signals
5. Update API URLs: Replace /v1/holographic/* with /v1/graph/* for standalone transforms/reranks
6. Test increased limits: Verify your app handles up to 200 search results

General Migrations:

1. Replace deprecated request fields in all write/search calls.
2. Add explicit tenant scope (organization_id, namespace_id) where needed.
3. Add explicit policy controls (memory_policy) for sensitive data.
4. Re-test retrieval with agentic graph enabled.
5. Update examples and SDK wrappers to emit only current fields.

Safe Baseline for New Integrations

{
  "external_user_id": "user_123",
  "memory_policy": {
    "mode": "auto",
    "consent": "implicit",
    "risk": "none"
  }
}

Then add manual mode and ACL specificity as your use case matures.

Validation Checklist

Search Ranking (Latest):

• ✅ No references to policy.vector in search calls
• ✅ Using reranking_config with correct provider names
• ✅ No references to deprecated rank_results
• ✅ No references to deprecated holographic_config in search

Graph-Aware Embeddings:

• ✅ No references to enable_holographic in add/update calls
• ✅ No references to frequency_schema_id (use domain_id instead)
• ✅ All /v1/holographic/* endpoints replaced with /v1/graph/*
• ✅ Domain registrations use domain_id and signals fields
• ✅ Search limits updated to support up to 200 results

General:

• ✅ Deprecated fields are removed from write and search payloads
• ✅ Current fields (external_user_id, memory_policy) are consistently used
• ✅ Integration tests pass using current schema contracts

Troubleshooting

If migration breaks existing behavior, compare payload diffs against Request Templates and validate with Error Playbook.