Skip to content

cianfhoghlaim/cianchosaint / .agents/skills/graphiti

graphiti

Verified

Graphiti - Temporal Knowledge Graph Memory

Version: graphiti-core 0.29.2+ | Last Updated: 2026-09-26 Verified upstream: https://github.com/getzep/graphiti/releases v0.29.2 (ff7e29c) Docs: https://help.getzep.com/graphiti (root URL — no double /graphiti/graphiti/ prefix) llms.txt: https://help.getzep.com/llms.txt (best for agents)

URLs (verified 2026-06-29)

Quick start

from graphiti_core import Graphiti
from graphiti_core.nodes import EpisodeType   # canonical since v0.29.x

# Canonical URI: falkor://host:port  (Cloud: falkor://your-instance.falkordb.cloud:6379)
graphiti = Graphiti(uri="falkor://falkordb:6379")
await graphiti.build_indices_and_constraints()  # one-time

# Add an episode (event-time + ingestion-time tracked)
await graphiti.add_episode(
    name="meeting-2026-06-29",
    episode_body="Cian met with the Bunchloch team to plan a new FalkorDB-backed curriculum graph.",
    source=EpisodeType.text,
    source_description="calendar entry",
    reference_time=datetime(2026, 6, 29, 10, 0),
)

# Search (bi-temporal + hybrid)
results = await graphiti.search_("FalkorDB curriculum", num_results=10)

v0.29.x new surface (live, not in Wave 1)

SymbolFirst releasedPurpose
Graphiti.summarize_saga(saga_id)v0.29.0First-class multi-episode narrative rollup. Backed by SagaNode with bi-temporal watermarks. Returns SagaNode.
EpisodeType.fact_triplev0.29.0Direct fact-triple ingestion (no LLM extraction step).
EpisodicNode.episode_metadata: dictv0.29.0Customer-defined filtering key.
_extract_and_resolve_nodes(episodes: list[EpisodicNode])v0.29.0Multi-episode batched node extraction; sets episode_indices on each node.
extract_edges(episodes: list[EpisodicNode])v0.29.0Multi-episode batched edge extraction (concatenates with [Episode N] headers).
extract_timestamps / extract_timestamps_batchv0.29.0Decoupled valid_at / invalid_at post-extraction step.
FalkorDriver(falkor_db=AsyncFalkorDB(...))v0.29.2Embedded FalkorDB Lite injection.
graphiti-core[falkordblite] extrav0.29.2pip install graphiti-core[falkordblite] for embedded mode (Python 3.12+).
OpenAIGenericClient default structured_output_mode="json_schema" w/ json_object fallbackv0.29.2 (#1537)Pragmatic structured-output for OpenAI-compatible endpoints.

Anti-patterns (D-series, new in v0.29.x)

  1. Don't pin to graphiti-core < 0.28.2 if you also use the MCP server. The mcp-v1.0.2 advisory is verbatim: "MCP v1.0.1 and prior has a security cypher injection vulnerability via graphiti-core 0.28.1." Use mcp >= 1.0.2 + graphiti-core >= 0.28.2.
  2. Don't parse [Episode N] headers assuming 1-based indexing. v0.29.0 switched to 0-based. Update episode_indices consumers accordingly.
  3. Don't import ops modules directly (e.g. from graphiti_core.driver.falkordb.operations.entity_edge_ops import FalkorEntityEdgeOperations). v0.28.0 redesigned the architecture; ops are injected via FalkorDriver._entity_edge_ops property. Direct import will break.
  4. Don't assume FalkorDB Lite is unavailable. v0.29.2 added graphiti-core[falkordblite] extra (Python 3.12+). For embedded mode pass FalkorDriver(falkor_db=AsyncFalkorDB(dbfilename=...)).
  5. Don't hard-code gpt-4o-mini as the default. v0.29.2 #1551 promoted gpt-5.5 as the upstream default with model-tied reasoning effort. Override with OpenAIClient(config=LLMConfig(model="your-model")).
  6. Don't use EpisodeType.text for Cianfhoghlaim product/catalog updates — EpisodeType.json bypasses entity extraction prompts and is ~10× faster.
  7. Don't assume the docs live at help.getzep.com/graphiti/graphiti/.... The Cianfhoghlaim .env.example should set GRAPHPITI_DOCS_URL=https://help.getzep.com/graphiti (root, no double prefix) for the OpenSpec "live-docs" verification.

Saga watermarks

# Both last_summarized_at (wall-clock filter) and
# last_summarized_episode_valid_at (temporal public watermark)
add_saga_watermarks: True

Live-docs verification (Agent 81 — 2026-06-29)

Wave 2 verified graphiti-core 0.29.2 is still the latest on PyPI & GitHub as of 2026-06-29. No release between Wave 1 (2026-06-28) and Wave 2 (2026-06-29). Next scheduled re-verify: when 0.29.3 tags appear on github.com/getzep/graphiti/releases or when 0.30.0 ships. Bookmark https://github.com/getzep/graphiti/releases.atom for RSS watching.

CCC anchors (existing + new)

  • graphiti_core/graphiti.py:980 — add_episode definition (line verified)
  • graphiti_core/driver/falkordb_driver.py:124 — FalkorDriver class
  • graphiti_core/driver/falkordb_driver.py:113 — default_group_id: str = '_' (new in v0.29.x)
  • graphiti_core/nodes.py — EpisodeType canonical location
  • graphiti_core/search/search_config_recipes.py — search recipes (NODE_HYBRID_SEARCH_RRF, etc.)
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall