RAG, embeddings, hallucination, PHI redaction, UPL safeguards, Quiz Mode, vertical specialisation, multi-tenant — every term grounded in a concrete SLAtech example. Linked from the FAQ and the comparison matrix.
A software agent that interprets natural-language input from a visitor and responds in kind, typically via a web widget, messaging channel (WhatsApp, Telegram), or voice interface. Modern AI chatbots are powered by large language models (LLMs) and retrieval-augmented generation (RAG); legacy chatbots used keyword-trigger flow editors.
HIPAA-required contract between a covered entity (clinic, hospital) and a business associate (vendor processing PHI). Without BAA, the firm carries direct liability for any PHI processed by the vendor. SLAtech Medical ships a BAA-eligible single-tenant option; BAA executed before any PHI ingest. "We're working on it" forever = red flag.
Bidirectional-text handling for Hebrew, Arabic, Persian — requires explicit awareness in form layouts, date pickers, currency strings, and button alignment. Most chatbot vendors auto-translate but don't bidi-localise. SLAtech ships RTL Hebrew first-class — bidi-aware UI primitives, locale-aware date format dd/mm/yyyy, ILS currency, Hebrew RFC2822 phone formatting.
Two-way integration with Google Calendar / Outlook so the bot can read live availability, book slots, send reminders and handle cancellations. SLAtech bundles calendar sync in the Pro tier ($249/month) — most competitors (Drift, Crisp) charge calendar booking separately.
Where the bot meets the visitor — web widget, Telegram, WhatsApp Business, Instagram DM, email. SLAtech bundles web + Telegram + WhatsApp in Pro. ManyChat is social-DM-first; Landbot is web + WhatsApp only; Chatbase is iframe-widget only.
A bite-size piece of indexed content (typically 200-500 tokens) that the retrieval pipeline returns as a citation source. Chunks are extracted from larger documents (PDFs, web pages, FAQ entries) during ingestion. SLAtech now ships a Snippet field per chunk in the bot response so widgets and AI scrapers can render the actual quoted text under each citation.
A reference shown alongside a bot answer pointing to the source content used to ground it. SLAtech bot responses ship { sourceUrl, snippet } per citation so the widget can render a "according to" hover-card and AI scrapers can extract grounded quotes.
A 0-1 number representing the LLM's certainty in an answer, derived from retrieval scores and LLM-as-Judge output. SLAtech surfaces confidence per response in the admin Inbox. Below 0.5 typically triggers a human-handoff fallback rather than a guessed answer.
The percentage of incoming conversations a bot resolves end-to-end without a human handoff. Containment-rate drives ROI directly — a 50% containment vs 80% containment is a 30-percentage-point cost-of-support difference. SLAtech baseline containment: Med 78%, Legal 73%, General-Business 72%. Tracked per-tenant in the admin dashboard.
The maximum number of tokens an LLM can process in a single request (system prompt + history + retrieval + question). Leading large language models ship a 128k token window. SLAtech truncates conversation history to the most recent N message pairs to stay below the limit.
A measure of semantic closeness between two embeddings, valued in [0, 1] (1 = identical). SLAtech's retrieval pipeline filters chunks by a configurable ScoreThreshold (default 0.5) and returns the top-K matches (default 10). Below threshold the bot routes to a human-handoff fallback.
Customer data physically stored within a specified jurisdiction, processed by sub-processors in that region, and subject to that jurisdiction's cross-border transfer rules. After the 2025 Schrems-III decision, cross-border transfer can be a compliance gate for regulated buyers. SLAtech offers hosting in your own jurisdiction on request (Enterprise).
Fee a vendor charges for exporting data out of their platform. Common lock-in vector — published low subscription price + hidden egress fee that triples cost-of-switching. SLAtech charges zero egress fee; documented in pricing page and in MSA. Ask any vendor "will you charge me for switching off?" — get the answer in writing.
A high-dimensional numeric vector that represents the semantic meaning of a text chunk. Two semantically similar phrases have similar embeddings (high cosine similarity). Embeddings power the retrieval half of RAG. SLAtech generates 1536-dimension embeddings with a leading embedding model, stored in its semantic vector store.
A sealed benchmark of N questions used to score a bot's responses across factuality, hallucination, and confidence axes. SLAtech ships a 200-question per-vertical eval harness — same questions used to compare SLAtech Medical vs SLAtech General-Business vs every competitor on identical input. Methodology is open-sourced at /en/eval/ so customers can reproduce scores against their own tenant.
US law protecting student-education records. Restricts disclosure of GPA, financial-aid status, disciplinary records without identity-verification. SLAtech Edu ships a FERPA-aware identity-verification step before disclosing protected records; generic chatbots over-disclose on impersonation attempts (compliance liability).
HL7 standard for exchanging healthcare data via REST APIs. Resources include Patient, Appointment, Practitioner, Encounter, Observation. SLAtech Medical can query FHIR-conformant EMRs to quote live appointment availability and practitioner-specific intake forms — not possible with generic chatbots that have no schema awareness.
Training an LLM on a tenant-specific dataset to specialise its tone or knowledge. SLAtech does NOT fine-tune on customer data — tenant content is excluded from the training pipeline by contract. RAG (not fine-tuning) handles tenant-specific knowledge, which is faster to update and avoids data-residency complications.
General Data Protection Regulation (EU 2016/679). SLAtech is GDPR-compliant by default: sub-processors governed by SCC 2021/914, no customer data used to train models, Data Subject Request portal in the admin platform. Hosting in your own jurisdiction is available on request (Enterprise).
When an LLM generates content that sounds plausible but isn't supported by the retrieved context or factually correct. Hallucinations are most dangerous in regulated verticals (clinical advice, legal positions, pricing). SLAtech surfaces a per-response hallucination flag in the admin Inbox so practitioners can intervene before a mis-routed answer becomes a complaint.
The corpus of tenant content (FAQ entries, PDFs, scraped web pages, manually-curated articles) that grounds bot answers via RAG. SLAtech's ingestion pipeline runs document extraction, chunking, embedding and vector-store upsert as a Worker job; reconciliation runs nightly.
A pattern where one LLM evaluates the output of another LLM on factuality, hallucination and confidence. SLAtech runs LLM-as-Judge on every bot response and surfaces the three scores in the admin Inbox so practitioners can audit quality without reading every transcript.
The bot workflow that collects visitor name + email + intent and pushes to a CRM. SLAtech supports HubSpot, Salesforce, Pipedrive natively and any in-house CRM via generic JSON-POST webhook. Lead-capture conversion is a primary success metric in the admin dashboard.
Sales-qualification frameworks. MEDDIC = Metrics, Economic-buyer, Decision-criteria, Decision-process, Identify-pain, Champion. BANT = Budget, Authority, Need, Timeline. SLAtech Sales scores leads on all six MEDDIC axes during intake. Generic chatbots collect name + email only — losing 80% of the qualification signal that MEDDIC captures.
Architecture pattern where a single deployment serves multiple paying tenants (customers), each with isolated data partitions. SLAtech is multi-tenant at every layer: DB tables, vector-store collections, Sentry projects, audit logs — all keyed by ClientId. No data is ever returned cross-tenant.
Protected Health Information and Personally Identifying Information — categories regulated under HIPAA (US), GDPR (EU) and similar frameworks. Across all verticals, SLAtech tokenises detected structured identifiers (email, phone, payment card, Israeli national-ID, medical-record number) in opt-in document uploads and redacts them from visitor messages before any LLM call; the originals stay encrypted at rest and free-text names are not auto-detected.
SLAtech Education feature that converts each uploaded lesson into a randomised practice quiz. Drives both the live Q&A bot and the per-topic self-test loop. Students can't game it by memorising — each quiz generates fresh with randomised order and phrasing.
An LLM grounding technique: before generating an answer, the system retrieves the most relevant chunks from a vector store seeded with tenant content, then passes those chunks as context to the LLM. Eliminates the LLM's reliance on training-data knowledge — answers are grounded in clinic / hotel / firm content and therefore won't hallucinate pricing or policy. SLAtech's bot pipeline is RAG-native.
Self-hosted = customer runs the bot on their own infrastructure (Botpress 11 ships open-source self-host). Cloud = vendor runs everything. SLAtech is cloud-only by default; Enterprise tier offers a single-tenant deployment option that mirrors self-host benefits without the ops burden.
Server-Sent Events transport that delivers LLM output token-by-token as it generates. Cuts perceived latency by ~70% vs waiting for the full response. SLAtech's /v1/query/ask-stream endpoint emits SSE with the final 'done' event carrying sources + snippet metadata.
A third party that processes customer data on behalf of SLAtech (an LLM provider for inference, SendGrid for email, Sentry for observability, Cloudflare for CDN). The public sub-processor list lives at /en/trust/.
The instruction text prepended to every LLM call, defining role, tone, constraints (e.g. "You are SLAtech Medical's intake assistant. Never offer clinical diagnoses"). SLAtech ships a per-vertical default system prompt that customers can override in the admin platform.
The total cost of running a chatbot deployment over its lifetime — subscription + residual human-support + integration cost + switching cost. SLAtech ships an interactive TCO calculator at /en/tco-calculator/ where you input monthly conversations, agent cost, handle-time, and containment-rate to get annual-savings + payback-period numbers.
An LLM sampling parameter that controls output randomness (0 = deterministic, 1 = creative, 2 = wild). SLAtech defaults to 0.3 for grounded customer-facing answers and 0.7 for creative replies in low-stakes channels. Low temperature reduces hallucination risk.
The unit LLMs operate on — roughly 4 chars or ¾ of an English word. Pricing is per-token (input + output). SLAtech reports per-query token usage and cost (EstimatedCostUsd) in the API response so customers can monitor spend in real time.
The retrieval-pipeline parameter that controls how many chunks are returned per query. SLAtech defaults to TopK=10, clamped to [1, 20]. Higher TopK gives the LLM more context but increases latency and cost; the sweet spot for most verticals is 5-10.
Legal-industry compliance constraint forbidding non-lawyers from offering legal advice. SLAtech Legal's UPL safeguard routes every substantive legal question to "an attorney will follow up" rather than letting the bot answer — preventing inadvertent UPL exposure for the firm.
Database optimised for cosine-similarity search across millions of embeddings. SLAtech runs its vector store under each tenant's ClientId, with strict multi-tenant isolation enforced at the partition key level. Every query is scoped to a single ClientId — no cross-tenant data is ever retrievable.
Contractual or technical constraints that make switching vendors prohibitively expensive — proprietary export formats, exit fees, mandatory implementation-consultant SOWs, knowledge-base schema lock. SLAtech ships zero egress fee, JSON-Lines conversation export with 24-hour SLA, and Markdown knowledge-base export. Documented in MSA — not buried in a "contact sales" gate.
Architectural pattern of shipping multiple industry-tuned bots (clinical, hospitality, legal, etc.) rather than one generic chatbot. SLAtech ships nine verticals — each with industry-specific FAQ templates, tone-of-voice presets and compliance posture seeded on signup.
No matching terms.
Four self-serve buyer-evaluation tools where the glossary terms above are operationalised:
Email the founder — we add terms as buyers ask them.