Skip to main content
POST
/
api
/
v1
/
memories
/
search
Python
from everos import EverOS

client = EverOS()
memories = client.v1.memories

response = memories.search(
    filters={"user_id": "<string>"},
    query="<string>",
    method="<string>",
    top_k=5
)
print(response)
{ "data": { "query": { "text": "<string>", "method": "<string>", "filters_applied": {} }, "episodes": [], "profiles": [], "raw_messages": [], "agent_memory": { "cases": [], "skills": [] }, "original_data": {} } }

Documentation Index

Fetch the complete documentation index at: https://docs.evermind.ai/llms.txt

Use this file to discover all available pages before exploring further.

Retrieval Methods
  • keyword: BM25 keyword retrieval (Elasticsearch only)
  • vector: Vector semantic retrieval (Milvus only)
  • hybrid (default): Hybrid retrieval with rerank (ES + Milvus + Rerank)
  • agentic: LLM-guided multi-round intelligent retrieval
Memory Types
  • episodic_memory: Episodic memory (ES + Milvus)
  • profile: User profile (Milvus only)
  • raw_message: Raw unprocessed messages (ES only)
  • agent_memory: Agent memory - cases and skills (ES + Milvus)
Filters DSL Same syntax as the GET endpoint. Must contain at least one of user_id or group_id.

Authorizations

​
Authorization
string
header
default:Bearer <api_key>
required

Bearer authentication header of the form Bearer 'api_key', obtain your API key from everos.evermind.ai.

Body

application/json
​
query
string
required

Search query text

Example:

"What did Alice say about the project?"

​
filters
object
required

Filter conditions. Supported fields: user_id, group_id, session_id, timestamp. user_id and group_id are placed at the top level of the filters object. session_id and timestamp support operators (eq, in, gt, gte, lt, lte) and can be used inside AND/OR combinators.

​
method
enum<string>
default:hybrid

Retrieval method

Available options:
keyword,
vector,
hybrid,
agentic
​
memory_types
enum<string>[]

Memory types to search

Available options:
episodic_memory,
profile,
raw_message,
agent_memory
​
top_k
integer
default:-1

Max results. -1 = return all meeting threshold (up to 100)

Required range: -1 <= x <= 100
​
radius
number | null

COSINE similarity threshold (0.0-1.0) for vector methods

Required range: 0 <= x <= 1
​
include_original_data
boolean
default:false

Whether to return original data

Response

Search results returned

​
data
object