aleth-client

A thin, standalone Python client for Aleth's persistent memory + thesis tracking over the warm-daemon wire. Part of MEM-SDK (ROADMAP ch.161, Option A). It imports no engine package and depends only on pydantic -- a sibling project (a Hand, another venture's agent) can pip install it and get durable memory without pulling in the Brain.

License (S1421): MIT. The client and the MPB-2 host plugins (clients/aleth-hermes, clients/aleth-openclaw) are open-source; the Aleth engine (the daemon this client talks to) stays closed.

Install

pip install ./clients/aleth-client

(In-repo install only -- no PyPI publish, by design. Copy the directory into a sibling project and pip install ./aleth-client there.)

Prerequisite: a running engine

The client talks to a running engine discovered via its .aleth_daemon.json discovery file (under data/runtime/ by default, or pass discovery_dir= / set ALETH_DISCOVERY_DIR). The daemon holds the warm graph; the client stays thin and never loads a local copy.

# the host that serves this wire in-repo (it writes .aleth_daemon.json):
python -m enki.runtime --pool-id baseline-pool --is-master --data-dir data --detach

Usage

from aleth_client import MemoryClient

mem = MemoryClient().connect()          # verifies reachability + protocol MAJOR

node = mem.remember("The sky is blue", origin="user")
print(node.node_id, node.classification)   # a computed judgment, not a constant

# Declare what you know about your own memory; your word outranks analysis.
fact = mem.remember("My locker code is 4312", classification="fact")

hits = mem.recall("sky", limit=5)
for n in hits.nodes:
    print(n.node_id, round(n.score, 3), n.content[:60])

# Thesis tracking
mem.store_reasoning_chain(
    chain_id="my_decision", title="My decision",
    steps=["weighed A [0.8]", "B cheaper [0.9]"], session_number=1,
)
h = mem.create_hypothesis(
    prediction="X holds", rationale="...", falsification_condition="...",
    confidence=0.6, hypothesis_type="completion", testability=0.8,
    testable_criteria="...", source_node_ids=[],
)
mem.add_hypothesis_evidence(
    hypothesis_id=h.node_id, outcome="confirmed",
    evidence_node_ids=[], confidence_change=0.1, experiment_id="exp1",
)

Method surface

Memory: remember, recall, recall_deep, recall_ranked, show, search, retract, connect_nodes, status, nodes. Thesis: store_intuition, store_reasoning_chain, session_checkpoint, recall_chains, recall_intuitions, create_hypothesis, add_hypothesis_evidence, retract_hypothesis. Goals: goal_create, goals_list, goal_detail, goal_progress, goal_achieve, goal_abandon. Governance: bios_check, bios_status. (No forget -- the memory surface is retract-based, matching the wire.)

Errors

DaemonUnavailable (no daemon answered), BackendError (daemon-side failure, incl. an [UNCERTAIN] write whose confirmation never arrived -- verify before retrying), NodeNotFound, BIOSDenied (blocked by a BIOS axiom), ProtocolMismatch (the daemon advertises a different contract MAJOR).

Versioning & compatibility

The wire contract is frozen at MEMSDK_PROTOCOL_VERSION = "2.3" and governed by semver (MAJOR = breaking, MINOR = additive, PATCH = fix). v2.3 (RV2-2, S1567) adds VALID TIME: the optional observed_at / valid_from / valid_to kwargs on remember and as_of on recall, plus valid_at and the three dates on each returned row, as_of / order on the result, and the three dates on show. A superseded belief becomes reachable BY DATE instead of only by being the newest. v2.2 (S1566) adds match_strength per row and the coverage verdict. v2.1 (EPI-1, S1565) adds the optional classification / confidence declaration kwargs on remember -- forwarded only when given, so older servers never see them. (This line read "2.1" while the constant was already "2.2"; corrected S1567 -- the number here is a CLAIM and must move with version.py.) v1.1 (MPB-2, S1421) adds recall_ranked -- daemon-side cross-encoder-ranked recall (bge-reranker-v2-m3 on CUDA hosts / ms-marco-MiniLM-L-12-v2 on CPU) whose nodes carry created_at for the dated presentation helper (aleth_client.presentation.dated_block). The daemon advertises its version via sdk_status; connect() checks the MAJOR matches and raises ProtocolMismatch otherwise. The full spec is docs/architecture/deployment/mem-sdk-contract.md; drift between this copy and the engine is caught by the conformance guard (tests/test_contract_conformance.py, MSDK-3).

Trust model (read this -- MSDK-5)

This client authenticates to the daemon with a local discovery-file token and inherits the same-box trust boundary: everything on the box is the operator. It does NOT provide per-consumer identity, tenancy isolation, or off-box transport, and a plain recall sees the GLOBAL memory graph (the source_project tag is a filter, not a partition). Do not expose this client across a trust boundary or treat it as an externally-safe, multi-tenant API. Anyone needing authenticated per-consumer identity, real tenancy, or an off-box endpoint wants the (un-built, rejected-for-now) Option B in docs/ROADMAP_ALT.md.

Which surface do I want?

• This client (aleth-client) -- you are writing Python code in a sibling project that needs durable memory over the wire.
• The foreign-MCP-window path -- you are an assistant window (Claude Code / VS Code) working inside another venture's repo and want the memory graph as MCP tools with project provenance. See docs/reference/using-enki-from-other-projects.md.