Install Conduit

Conduit installs as a local CLI plus a background daemon that manages your knowledge base, containerized services, and health checks. After installation, you can ingest sources, run RAG (vector search) immediately, and opt into KAG (knowledge graph + multi-hop reasoning) only when it makes sense for your domain. The goal is fast local setup without wiring a dozen services by hand.

Image

conduit-install-overview | ALT: "Conduit installation and local runtime overview (CLI + local services)"

Caption: Conduit runs locally with a CLI, daemon, and containerized services for search and graph workloads.

Installation and Setup (step-by-step)

Step 0: prerequisites

macOS or Linux (Conduit runs as a daemon: launchd on macOS, systemd on Linux).
A container runtime: Podman (preferred) or Docker (supported).
Git (used by the installer when building from source).

The installer can install missing dependencies (Go, Git, Podman/Docker, Ollama), but you still need a container runtime available for Qdrant and FalkorDB containers.

Step 1: install Conduit

curl -fsSL https://raw.githubusercontent.com/amlandas/Conduit-AI-Intelligence-Hub/main/scripts/install.sh | bash

Step 2: verify the install

conduit --version
conduit status
conduit doctor

Step 3: start/verify services

The daemon starts automatically. If it is not running, start it and confirm status:

conduit service start
conduit service status
conduit status

If you need to install or re-install the local services explicitly:

conduit qdrant install
conduit falkordb install

System requirements (what you need installed)

Container runtime: Podman (preferred for rootless performance) or Docker (supported).
Qdrant: vector database used by RAG. Conduit can install/manage it via conduit qdrant install.
FalkorDB: graph database for KAG. Conduit can install/manage it via conduit falkordb install.
Ollama: used for embeddings and KAG extraction. The installer can install it and download a default model.
Git: required for installation/build steps.

If you build from source, the docs specify these build/runtime dependencies:

Go 1.21+ (build)
SQLite 3.35+ (runtime)
GCC (CGO build support)

Experimental GUI: Conduit ships an experimental GUI for macOS. Keep the install focused on CLI setup and review the GUI status here: /docs/gui.

Choosing RAG vs KAG (when to use which)

RAG (default): fast vector search using Qdrant. This is the right choice for most workflows and scales well with frequent syncs.
KAG (optional): builds a domain knowledge graph and supports multi-hop reasoning. It is disabled by default and takes longer to build depending on knowledge base size.

When KAG is worth it

KAG is most valuable when:

You have a stable ontology or strong domain structure.
The domain is naturally expressed as entities + relations.
Queries require multi-hop, constraint-heavy reasoning where paths matter.
You need high accuracy, explainability, and auditability.
Workloads are entity-centric (facts, systems, inventories), not open-ended chat.
The upfront graph cost is amortized across many workflows and teams.

Optional: enable Knowledge-Augmented Generation (KAG)

KAG builds a graph from your KB and can take a while depending on corpus size. Make sure FalkorDB is available, then sync:

conduit falkordb install
conduit falkordb start
conduit kb kag-sync

Example KAG queries:

conduit kb kag-sync
conduit kb kag-query "Kubernetes"
conduit kb kag-query "authentication" --entities OAuth,JWT --max-hops 2

If KAG feels slow or too expensive for your use case, stick with RAG and revisit later.

Uninstall

Recommended (interactive):

conduit uninstall

If the CLI is unavailable, use the uninstall script:

curl -fsSL https://raw.githubusercontent.com/amlandas/Conduit-AI-Intelligence-Hub/main/scripts/uninstall.sh | bash

To remove data as well (this deletes your local index and KB state):

curl -fsSL https://raw.githubusercontent.com/amlandas/Conduit-AI-Intelligence-Hub/main/scripts/uninstall.sh | bash -s -- --remove-data

The uninstall script does not remove dependencies (Ollama, Podman/Docker, containers). If you want to remove Qdrant/FalkorDB containers, use your container runtime to remove them after uninstall.

Common issues

Qdrant missing / 0 vectors

Detect the issue:

conduit status

Diagnose and recover:

conduit doctor
conduit qdrant stop
conduit qdrant start
conduit kb sync --rebuild-vectors

Container storage mount issues (first install)

Restart the affected container:

conduit qdrant stop
conduit qdrant start
conduit falkordb stop
conduit falkordb start

For deeper debugging, see /docs/troubleshooting and the upstream Known Issues list: https://github.com/amlandas/Conduit-AI-Intelligence-Hub/blob/main/docs/KNOWN_ISSUES.md