Memory Prefetch

Prefetch pulls relevant long-term memories into the system prompt at the start of every turn — the agent answers with context it never had to ask for.

Quick Start

Enable with three lines

Set prefetch=True on MemoryConfig and the agent recalls relevant memories before its first model call.

from praisonaiagents import Agent, MemoryConfig

agent = Agent(
    name="assistant",
    instructions="Answer using recalled memories when relevant.",
    memory=MemoryConfig(
        user_id="alice",
        prefetch=True,   # that's it
    ),
)

agent.start("What did I say about my preferences?")

Tune the budget

Cap how many memories are injected and how large the block can grow.

from praisonaiagents import Agent, MemoryConfig

agent = Agent(
    name="assistant",
    instructions="Answer using recalled memories when relevant.",
    memory=MemoryConfig(
        user_id="alice",
        prefetch=True,
        prefetch_limit=3,           # inject at most 3 memories
        prefetch_token_budget=200,  # estimated-token cap; truncates with "…"
    ),
)

agent.start("What are my UI preferences?")

Cross-agent recall

Store facts with one agent and recall them in another by sharing the same backend and user_id.

from praisonaiagents import Agent, MemoryConfig

shared = MemoryConfig(backend="sqlite", user_id="alice")

writer = Agent(name="writer", instructions="Store facts.", memory=shared)
writer.memory.remember("Launch date is Nov 15.")

reader = Agent(
    name="reader",
    instructions="Answer using recalled context.",
    memory=MemoryConfig(**{**shared.to_dict(), "prefetch": True}),
)
print(reader.start("When do we launch?"))

How It Works

Prefetch searches long-term memory with the user prompt, dedupes and budgets the matches, then appends them to the system prompt before the first LLM call.

Phase	What happens
1. Query	The user prompt drives a `search_long_term` call, scoped by `user_id` / `session_id`
2. Shape	Matches are deduplicated by text, then truncated to the token budget with a trailing `…`
3. Inject	The result is appended under `## Recalled memories` in the system prompt

Prefetch is default-off. When prefetch is False (or there is no memory instance), the backend is never queried and the prompt is unchanged. Backend errors are swallowed at debug level — the turn continues without recalled context.

Configuration Options

Full list of options, types, and defaults — MemoryConfig

Option	Type	Default	Description
`prefetch`	`bool`	`False`	Master switch. Off = zero backend calls.
`prefetch_limit`	`int`	`5`	Max memories injected per turn.
`prefetch_token_budget`	`int`	`512`	Estimated-token cap for the injected block. Overflow is truncated with `…`.

Identity scoping (user_id, session_id) is inherited from the same MemoryConfig.

Common Patterns

Pattern 1 — Personal assistant with recalled preferences

from praisonaiagents import Agent, MemoryConfig

agent = Agent(
    name="assistant",
    instructions="Answer using recalled preferences.",
    memory=MemoryConfig(user_id="alice", prefetch=True),
)
agent.memory.remember("Alice prefers metric units and dark mode.")
print(agent.start("How should I format my dashboard?"))

Pattern 2 — Tight budget for cost control

from praisonaiagents import Agent, MemoryConfig

agent = Agent(
    name="assistant",
    instructions="Answer concisely using recalled context.",
    memory=MemoryConfig(
        user_id="alice",
        prefetch=True,
        prefetch_limit=2,
        prefetch_token_budget=128,
    ),
)
print(agent.start("Remind me what I asked for last time."))

Best Practices

Default-off is intentional

Enable prefetch per agent, not globally, so you pay the backend query only on the turns that benefit from recalled context.

Pair with user_id

Without user_id, prefetch searches the per-instance store — a fresh agent-<uuid> on every process. Set user_id so recalled memories persist across runs.

Keep the token budget small

prefetch_token_budget=512 is the ceiling, not the target. Large recalled blocks push out user context — start small and raise only when needed.

Async works out of the box

The async path awaits async backends; sync backends are offloaded off the event loop. No extra configuration required.

Backend failures never break the turn

Any backend error is logged at debug level and the agent proceeds without recalled context — prefetch never fails a turn.

Memory

Memory backends and the remember/recall API

Knowledge

Add documents and URLs as agent knowledge

Getting Started

Install

Core Concepts

Guides

Features

Models

Databases

Observability

Memory

Telemetry & Analytics

Knowledge

RAG

Persistence

Tools

Other Features

Developers

Configuration

Best Practices

Getting Started (No Code)

Quick Start

How It Works

Configuration Options

Common Patterns

Pattern 1 — Personal assistant with recalled preferences

Pattern 2 — Tight budget for cost control

Best Practices

Memory

Knowledge

​Quick Start

​How It Works

​Configuration Options

​Common Patterns

​Pattern 1 — Personal assistant with recalled preferences

​Pattern 2 — Tight budget for cost control

​Best Practices

​Related

Memory

Knowledge

Quick Start

How It Works

Configuration Options

Common Patterns

Pattern 1 — Personal assistant with recalled preferences

Pattern 2 — Tight budget for cost control

Best Practices

Related