The dbt Semantic Layer alternative with verified rules and recursion
SynaLemma is a semantic layer for AI agents: open-ended rules compiled to SQL, verified before they run, with recursion and assertions on the data.
SynaLemma vs dbt Semantic Layer at a glance
| Criterion | SynaLemma | dbt Semantic Layer |
|---|---|---|
| Definitions written in | Synalog, a logic language with a compiler | YAML on top of dbt models (MetricFlow) |
| What you can define | Any rule, built on other rules | Five metric types; no custom aggregations |
| Joins | As far as the question needs | At most two hops |
| Hierarchies and multi-hop chains | Native recursion, verified to terminate | No recursion |
| What checks the definitions | A symbolic verifier proves every rule well-formed, safe and terminating | Configuration checks, and the generated SQL must run |
| Checks on the data | Assertions in logic on any rule, with counterexamples | dbt data tests on models: keys, nulls, references |
| What an agent can run | Only the verified layer; a read-only account is enough | Metric queries, plus text-to-SQL and execute_sql, which does not block writes |
| Deployment | Your infrastructure, with the model you choose | The Semantic Layer APIs need a paid dbt platform plan |
Why teams look for a dbt Semantic Layer alternative
-
Five metric types. MetricFlow has no custom aggregations (metrics overview, FAQ).
-
Two-hop joins. MetricFlow joins at most three tables (join logic).
-
No recursion. Hierarchies and multi-hop chains cannot be modelled.
-
Configuration checks. Validation checks the configuration and that the SQL runs (validations).
-
Write-capable SQL for agents. The MCP server's execute_sql does not block write statements (MCP tools).
SynaLemma vs dbt Semantic Layer: key differences
1. Any rule, built on other rules
No fixed catalogue of metric types. Rules use other rules, with aggregation, negation and time.
2. Joins and recursion without limits
Joins go as far as the question needs. Recursive rules follow chains of any depth, verified to terminate.
3. Formal verification and assertions
Every rule is proved well-formed, safe and terminating. Assertions in first-order logic apply to any rule and return counterexamples.
4. One read-only path for agents
The analyst reaches your data only through the verified layer. A read-only database account is enough.
5. Sovereign by construction, not by contract
SynaLemma runs on-premises: data, rules and model stay on your infrastructure. The engine compiles the SQL, so a small open model is enough.
Example: a recursive query with assertions
A bill of materials at any depth, with two assertions. MetricFlow stops at two hops. The data contains a cycle; the engine reports it before any answer is returned.
# The bill of materials, read from the database.
Component(assembly:, part:, quantity:) :- bom(assembly:, part:, quantity:);
# A part is inside an assembly directly, or inside one of its parts.
@Recursive(Contains, 20);
@Assert(Contains,
transitive: "∀ a b c, Contains a b → Contains b c → Contains a c",
no_cycle: "∀ a, ¬ Contains a a");
Contains(assembly:, part:) distinct :- Component(assembly:, part:);
Contains(assembly:, part:) distinct :-
Contains(assembly:, part: sub),
Component(assembly: sub, part:);Sources
Checked against each product's public documentation on the review date. To report an error: contact@synalinks.com.
dbt Semantic Layer alternative: FAQ
What is the best dbt Semantic Layer alternative for AI agents?
SynaLemma. It has no fixed metric types or two-hop limit, verifies every rule, checks the data with assertions and gives agents a read-only path.
Can MetricFlow join more than two hops?
No. dbt's documentation sets a limit of two hops. In Synalog, rules join as far as the question needs.
Do I need dbt to use SynaLemma?
No. SynaLemma connects to your database directly and reads the tables your dbt models already build.
Can SynaLemma run with a local model?
Yes. A small open model on your own hardware is enough, because the engine compiles the SQL.
Bring a question your current stack can't answer.
In 30 minutes, we show SynaLemma answering it, with the rule behind the number and the checks it passed.