Skip to content

Glossary

This glossary defines the technical terms used across this documentation. Each entry answers the question “What is it?” with a concise definition followed by usage context inside the infrastructure.


Tool to define and run multi-container Docker applications. A docker-compose.yaml file describes the services, networks and volumes of a stack. A simpler alternative to Kubernetes for single-server deployments.

See: VPS Architecture

A set of Docker services that work together and share a network. Example: the n8n-stack contains N8N, Redis, PostgreSQL and the workers. Each stack has its own docker-compose.yaml.

Intermediary server that receives client requests and forwards them to the appropriate backend servers. Benefits: centralised TLS termination, domain-based routing, load balancing.

In this infrastructure: Caddy plays this role.

See: Security Stack

Encryption protocols that secure HTTP traffic (HTTPS). TLS is the successor to SSL. Let’s Encrypt provides free TLS certificates automatically through Caddy (ACME protocol).

Application-level firewall that filters and blocks malicious requests (SQL injection, XSS, known exploits). Works at the HTTP layer, unlike traditional network firewalls.

In this infrastructure: CrowdSec analyses logs and maintains a block list.

In CrowdSec, the component that enforces decisions. The Caddy bouncer rejects requests from banned IPs before they reach the backend services.

See: Security Stack

Periodic verification of a service’s state. Docker can automatically restart containers whose health check fails. Implemented through HEALTHCHECK commands in Dockerfiles or custom scripts.

See: Global health check


N8N execution mode where workflows are placed in a Redis queue (Bull Queue) and processed by separate workers. Enables horizontal scaling and resilience.

See: N8N in queue mode

FIFO (First In, First Out) data structure used to process tasks asynchronously. The producer adds tasks, the consumer processes them. Decouples emission from execution.

In this infrastructure: Redis stores Bull queues for N8N.

Process that consumes tasks from a queue. N8N workers run workflows in parallel to the main process, allowing several workflows to run at once.

Node.js library that uses Redis to manage job queues with automatic retries, priorities, delays and tracking events. N8N’s default queue backend.

HTTP URL that receives external event notifications via POST. Enables real-time integration between services without polling.

Examples: GitHub → N8N (issue created), Cal.com → N8N (booking created), Odoo → N8N (lead created).

N8N feature that stores structured data directly inside the instance, without an external database. Useful for small datasets (configuration, admin lists, cache).

N8N workflow called by another workflow through the “Execute Workflow” node. Enables reuse and modularisation. The sub-workflow receives input data and returns a result.

Example: Voice Transcription is a sub-workflow called by the Telegram Orchestrator.


Integrated software that centralises business data: projects, contacts, invoices, stock, HR, etc. Removes data silos and automates cross-department processes.

In this infrastructure: Odoo 18 Community.

See: Why Odoo

Remote procedure call protocol that uses XML over HTTP. Lets you call methods on a remote server as if they were local.

In this infrastructure: Odoo’s native API for integration with N8N and external systems.

Odoo extension that adds functionality. Installable through the UI or by adding the code into /mnt/extra-addons.

Custom addons: project_github_sync, website_experiments, TimeTrackr.

See: Odoo 18 Setup

Customer relationship management module. Tracks the lifecycle of prospects (leads): acquisition, qualification, proposal, negotiation, conversion.

Pipeline: Initial contact → Qualification → Proposal → Negotiation → Won/Lost.

In the Odoo CRM, a lead represents a potential prospect. It becomes an opportunity once qualified. The probability field captures the conversion likelihood (0–100%).


Database optimised to store and search vectors (embeddings) by similarity. Uses algorithms like HNSW for fast high-dimensional search.

In this infrastructure: Qdrant.

See: AI Stack

Numerical representation (vector of floats) of a piece of text, image or other data. Two semantically close items have close embeddings (low cosine distance).

Models: text-embedding-3-small (OpenAI), all-MiniLM-L6-v2 (sentence-transformers).

Technique that enriches LLM prompts with information retrieved from a knowledge base. Reduces hallucinations and lets the model use data it did not see during training.

Flow: Query → Embedding → Vector search → Context injection → LLM response.

Large language model trained on massive text corpora. Able to understand and generate text, answer questions, summarise, translate.

Examples: Claude (Anthropic), GPT-4 (OpenAI), Llama (Meta).

Execution mode without human approval. -yolo requests in Claude Ollama run directly without confirmation. Useful for trusted operations.

See: AI Stack

Speech-to-text model developed by OpenAI. Supports 99 languages and produces transcripts with optional timestamps.

Implementations: OpenAI API, Groq (free, fast), local (whisper.cpp).

See: Voice transcription

Identification and separation of speakers in an audio recording. Tells you “who said what” in a multi-participant conversation.

Service: ElevenLabs Scribe supports diarisation.


Monitoring system that collects metrics (scraping) from HTTP endpoints and lets you define alerts via PromQL. Pull model: Prometheus pulls the metrics.

See: Monitoring Stack

Prometheus query language. Lets you aggregate, filter and transform time-series metrics.

Example: rate(container_cpu_usage_seconds_total[5m]) computes the CPU usage over 5 minutes.

Visualisation platform that builds dashboards from data sources like Prometheus, InfluxDB, PostgreSQL. Supports alerts and annotations.

Prometheus component that handles alerts: grouping, temporary silencing, routing to different channels (webhook, email, Slack).

In this infrastructure: Routes alerts to the N8N Notification Hub.

Open-source standard for collecting metrics, logs and distributed traces. Decouples data generation from storage (vendor-neutral).

In this infrastructure: The OTEL Collector aggregates Claude Code data.

Prometheus action that periodically pulls metrics from /metrics endpoints. Typical interval: 15–60 seconds.

Service that exposes metrics in the Prometheus format (text/plain with labels). Bridges a system and Prometheus.

Examples: Node Exporter (system metrics), cAdvisor (Docker metrics), postgres_exporter.


Collaborative security solution that analyses logs to detect attacks and shares blocking decisions across worldwide instances. An open-source alternative to Fail2ban with collective intelligence.

See: Security Stack

In CrowdSec, a set of parsers (log analysis) and scenarios (pattern detection) used to detect specific attacks.

Examples: crowdsecurity/http-cve (known exploits), crowdsecurity/nginx-proxy-manager (NPM logs).

In CrowdSec, the action to take on a detected IP: ban, captcha, throttling. Decisions have a duration and can be shared with the community.

Free, automated certificate authority (ACME). Issues TLS certificates valid for 90 days, renewed automatically by Caddy.

Latest version of the HTTP protocol, built on UDP (QUIC) instead of TCP. Delivers better performance on unreliable networks (mobile, Wi-Fi) thanks to the removal of head-of-line blocking.

HMAC (Hash-based Message Authentication Code)

Section titled “HMAC (Hash-based Message Authentication Code)”

Integrity and authenticity check that uses a shared secret. Used to verify that webhooks come from the expected source (GitHub, Stripe).

See: GitHub-Odoo sync


Service that watches Docker registries (Docker Hub, GHCR, etc.) and signals when new image versions ship. Enables controlled updates.

See: Docker auto-updates

Simple push notification server, self-hosted or SaaS. Lets you send notifications via plain HTTP POST without an SDK. Supports public and private topics.

Architectural pattern centralising every notification inside a single workflow. Responsible for routing (channel), formatting (template), deduplication and quiet-hour compliance.

See: Notification Hub

Time windows during which non-critical notifications are suppressed or queued. Avoids waking up admins for non-urgent alerts.


Open-source alternative to Calendly for booking appointments. Lets clients reserve slots based on configured availability.

See: Cal.com → Odoo CRM

Standard format (RFC 5545) for exchanging calendar events. An .ics file can be imported into Google Calendar, Outlook, Apple Calendar, etc.

Structure: VCALENDAR containing VEVENTs with DTSTART, DTEND, SUMMARY, etc.


Reference to another Git repository included inside a parent repo. Lets you version external code while keeping its own history. Requires git submodule update --init after a clone.

In this infrastructure: n8n-exports/ and docs-external/ are submodules.

Verification mechanism (typically HMAC-SHA256) that guarantees a webhook truly comes from the expected source and was not tampered with in transit.

See: GitHub-Odoo sync

Metadata file describing the content of an export or package. Usually contains the list of items, their versions, and the export date.

Example: manifest.json in N8N exports lists the workflows with their updatedAt.

See: N8N export to GitHub