Notes on governed semantic layers, from the people building datahashi.https://datahashi.com/https://datahashi.com/blog/2026-10-30-support-copilot-customer-health/https://datahashi.com/blog/2026-10-30-support-copilot-customer-health/How a support copilot answers an account-health question from certified metrics, with sample schema, model and queries that separate ticket volume from open critical escalations.Fri, 30 Oct 2026 00:00:00 GMTAgents & MCPagentscustomer-successmcpfilteringhttps://datahashi.com/blog/2026-10-29-self-hosted-regulated-agents/https://datahashi.com/blog/2026-10-29-self-hosted-regulated-agents/How a healthcare compliance team gave LLM agents access to clinical treatment analytics inside a private VPC without exposing database passwords.Thu, 29 Oct 2026 00:00:00 GMTAgents & MCPagentssecurityself-hostedmcphttps://datahashi.com/blog/2026-10-28-schema-change-without-breaking-dashboards/https://datahashi.com/blog/2026-10-28-schema-change-without-breaking-dashboards/Upstream dbt jobs renamed an invoice column to amount_usd, but central semantic mapping and model tests kept dashboards running without a blip.Wed, 28 Oct 2026 00:00:00 GMTGovernance & trustmigrationsschema-driftmodel-testslineagehttps://datahashi.com/blog/2026-10-27-marketing-cac-roas-match-finance/https://datahashi.com/blog/2026-10-27-marketing-cac-roas-match-finance/How governed derived metrics and transparent lineage unify customer acquisition cost and return on ad spend across marketing channels and finance.Tue, 27 Oct 2026 00:00:00 GMTModeling & metricsmetricsmarketinglineagecachttps://datahashi.com/blog/2026-10-26-rollups-cut-warehouse-bill/https://datahashi.com/blog/2026-10-26-rollups-cut-warehouse-bill/How pre-aggregation rollups serve morning executive dashboard queries from pre-computed tables rather than scanning raw transaction tables.Mon, 26 Oct 2026 00:00:00 GMTQueryingrollupspre-aggregationwarehouse-costsperformancehttps://datahashi.com/blog/2026-10-25-multi-tenant-customer-dashboards/https://datahashi.com/blog/2026-10-25-multi-tenant-customer-dashboards/How platform engineers secure embedded customer dashboards using JWT claims, row-level policies, and query metering in a thin semantic layer.Sun, 25 Oct 2026 00:00:00 GMTGovernance & trustmulti-tenantrlsjwtsecurityhttps://datahashi.com/blog/2026-10-24-ecommerce-category-tree-revenue/https://datahashi.com/blog/2026-10-24-ecommerce-category-tree-revenue/How retail merchandisers drill down SKU, category, and department revenue without fan-out bugs in a thin semantic layer.Sat, 24 Oct 2026 00:00:00 GMTModeling & metricshierarchiesfan-outecommercemerchandisinghttps://datahashi.com/blog/2026-10-23-saas-activation-retention/https://datahashi.com/blog/2026-10-23-saas-activation-retention/How product managers analyze SaaS user activation and weekly retention across entity join graphs in a thin semantic layer.Fri, 23 Oct 2026 00:00:00 GMTModeling & metricsretentionactivationentitiestime-grainshttps://datahashi.com/blog/2026-10-22-slack-bot-weekly-results/https://datahashi.com/blog/2026-10-22-slack-bot-weekly-results/How operations teams connect Slack bots to certified metrics using Model Context Protocol and query time zones in a thin semantic layer.Thu, 22 Oct 2026 00:00:00 GMTAgents & MCPmcpslack-bottime-zonesoperationshttps://datahashi.com/blog/2026-10-21-finance-close-one-arr/https://datahashi.com/blog/2026-10-21-finance-close-one-arr/How semi-additive snapshot measures and model tests keep ARR balances consistent across reporting tools in a thin semantic layer.Wed, 21 Oct 2026 00:00:00 GMTModeling & metricsarrsemi-additivemodel-testsfinancehttps://datahashi.com/blog/2026-10-10-bigquery-least-privilege-agents/https://datahashi.com/blog/2026-10-10-bigquery-least-privilege-agents/Apply BigQuery least privilege so agent-backed semantic serving never needs broad project roles or warehouse passwords in the agent.Sat, 10 Oct 2026 00:00:00 GMTWarehouses & setupbigquerywarehousehttps://datahashi.com/blog/2026-10-09-snowflake-readonly-semantic-layer/https://datahashi.com/blog/2026-10-09-snowflake-readonly-semantic-layer/A semantic layer on Snowflake is only as safe as the role that runs its queries. datahashi's expected posture is least-privilege, read-only roles for query execution.Fri, 09 Oct 2026 00:00:00 GMTWarehouses & setupsnowflakewarehousehttps://datahashi.com/blog/2026-10-08-duckdb-sandbox-first-metric/https://datahashi.com/blog/2026-10-08-duckdb-sandbox-first-metric/Stand up a DuckDB sandbox, publish one certified metric, and query it from the console and MCP without a cloud warehouse.Thu, 08 Oct 2026 00:00:00 GMTWarehouses & setupduckdbtutorialhttps://datahashi.com/blog/2026-10-07-eval-harness-same-number/https://datahashi.com/blog/2026-10-07-eval-harness-same-number/Create an eval harness that asserts agent answers match dashboard numbers for the same metric requests.Wed, 07 Oct 2026 00:00:00 GMTGovernance & trustagentsevalgovernancehttps://datahashi.com/blog/2026-10-06-mcp-auth-no-warehouse-passwords/https://datahashi.com/blog/2026-10-06-mcp-auth-no-warehouse-passwords/Set up MCP auth so agents never hold warehouse passwords - credentials stay server-side behind the semantic layer.Tue, 06 Oct 2026 00:00:00 GMTGovernance & trustmcpauthsecurityhttps://datahashi.com/blog/2026-10-05-grounding-copilots-governed-metrics/https://datahashi.com/blog/2026-10-05-grounding-copilots-governed-metrics/Ground copilots on certified metrics so natural-language answers inherit governance instead of table dumps.Mon, 05 Oct 2026 00:00:00 GMTAgents & MCPagentscopilotsmcphttps://datahashi.com/blog/2026-10-04-agent-wrong-numbers-five-modes/https://datahashi.com/blog/2026-10-04-agent-wrong-numbers-five-modes/A composite war story — five failure modes that show up when agents get put on metrics before the rails are ready, and how a governed path changes the ending.Sun, 04 Oct 2026 00:00:00 GMTAgents & MCPagentswar-storyhttps://datahashi.com/blog/2026-10-03-designing-safe-mcp-tools/https://datahashi.com/blog/2026-10-03-designing-safe-mcp-tools/MCP makes it easy to expose too much. datahashi treats MCP tool design as safety engineering — small verbs, strict schemas, fail-closed validation.Sat, 03 Oct 2026 00:00:00 GMTAgents & MCPmcpagentshttps://datahashi.com/blog/2026-10-02-connect-cursor-mcp-metrics/https://datahashi.com/blog/2026-10-02-connect-cursor-mcp-metrics/Connect Cursor to governed metrics over MCP — catalog discovery and typed queries for coding agents, without raw warehouse access.Fri, 02 Oct 2026 00:00:00 GMTAgents & MCPmcpcursorhttps://datahashi.com/blog/2026-10-01-mcp-without-raw-table-access/https://datahashi.com/blog/2026-10-01-mcp-without-raw-table-access/How agents can query governed metrics through MCP without seeing raw warehouse tables or writing SQL.Thu, 01 Oct 2026 00:00:00 GMTAgents & MCPmcpagentshttps://datahashi.com/blog/2026-09-30-prove-two-consumers-same-number/https://datahashi.com/blog/2026-09-30-prove-two-consumers-same-number/Run a practical same-number test between two consumers (dashboard and agent) against one metric request.Wed, 30 Sep 2026 00:00:00 GMTGovernance & trustgovernancetutorialagentshttps://datahashi.com/blog/2026-09-29-agent-safe-query-defaults/https://datahashi.com/blog/2026-09-29-agent-safe-query-defaults/Agents ask for everything by default. datahashi ships agent-safe limits so the first query an agent runs is already bounded, typed, and reviewable.Tue, 29 Sep 2026 00:00:00 GMTAgents & MCPagentsqueryinghttps://datahashi.com/blog/2026-09-28-reading-compiled-sql-debug/https://datahashi.com/blog/2026-09-28-reading-compiled-sql-debug/When a semantic query returns a number that feels wrong, read the compiled SQL. datahashi exposes it from `run_semantic_query` so you can see exactly how names became joins and filters.Mon, 28 Sep 2026 00:00:00 GMTQueryingqueryingdebugginghttps://datahashi.com/blog/2026-09-27-rollups-hot-questions/https://datahashi.com/blog/2026-09-27-rollups-hot-questions/Rollups now ship in datahashi, hosted and self-hosted. Declare one in YAML, and the engine builds it through your Query Runner and answers matching queries from it - with the limits stated up front.Sun, 27 Sep 2026 00:00:00 GMTQueryingqueryingperformancehttps://datahashi.com/blog/2026-09-26-filters-that-compose-safely/https://datahashi.com/blog/2026-09-26-filters-that-compose-safely/Two filters interacting should narrow a result, not rewrite the metric. How datahashi treats filter composition as a first-class part of every semantic query.Sat, 26 Sep 2026 00:00:00 GMTQueryingqueryingfiltershttps://datahashi.com/blog/2026-09-25-how-semantic-queries-become-sql/https://datahashi.com/blog/2026-09-25-how-semantic-queries-become-sql/A semantic query names metrics and dimensions; the warehouse still speaks SQL. Here's how datahashi compiles one into SQL you can inspect and trust.Fri, 25 Sep 2026 00:00:00 GMTQueryingqueryingsqlhttps://datahashi.com/blog/2026-09-24-semantic-query-101/https://datahashi.com/blog/2026-09-24-semantic-query-101/Introduce the semantic query triad so any consumer can ask the same question the same way.Thu, 24 Sep 2026 00:00:00 GMTQueryingqueryingtutorialsemantic-layerhttps://datahashi.com/blog/2026-09-23-derived-metrics-without-double-counting/https://datahashi.com/blog/2026-09-23-derived-metrics-without-double-counting/Derived metrics compose other metrics, not raw columns — get the expansion order wrong and a ratio silently double-counts. Here's how datahashi keeps composition honest.Wed, 23 Sep 2026 00:00:00 GMTModeling & metricsmetricsmodelinghttps://datahashi.com/blog/2026-09-22-entities-relationships-semantic-model/https://datahashi.com/blog/2026-09-22-entities-relationships-semantic-model/Metrics sit on entities. Relationships tell the layer how to join without letting every consumer reinvent the graph.Tue, 22 Sep 2026 00:00:00 GMTModeling & metricsmodelingsemantic-layerhttps://datahashi.com/blog/2026-09-21-time-grains-asymmetric-aggregates/https://datahashi.com/blog/2026-09-21-time-grains-asymmetric-aggregates/Time grain is part of the metric contract. Some aggregates roll cleanly across grains. Others do not, and agents need to know the difference.Mon, 21 Sep 2026 00:00:00 GMTModeling & metricsmetricstimehttps://datahashi.com/blog/2026-09-20-additive-vs-non-additive-metrics/https://datahashi.com/blog/2026-09-20-additive-vs-non-additive-metrics/Why some metrics can be summed across dimensions and time, and why others must never be.Sun, 20 Sep 2026 00:00:00 GMTModeling & metricsmetricsmodelinghttps://datahashi.com/blog/2026-09-19-fan-out-safe-metrics-explained/https://datahashi.com/blog/2026-09-19-fan-out-safe-metrics-explained/What fan-out risk looks like in metric queries and how to keep aggregates correct.Sat, 19 Sep 2026 00:00:00 GMTModeling & metricsmetricsmodelinghttps://datahashi.com/blog/2026-09-18-dimensions-without-exploding-joins/https://datahashi.com/blog/2026-09-18-dimensions-without-exploding-joins/How to model dimensions so metric grain stays correct and joins do not multiply rows.Fri, 18 Sep 2026 00:00:00 GMTModeling & metricsmetricsmodelinghttps://datahashi.com/blog/2026-09-17-first-metric-name-it/https://datahashi.com/blog/2026-09-17-first-metric-name-it/Walk through defining a first certified metric by name and measure without pasting hand-written SQL into the agent path.Thu, 17 Sep 2026 00:00:00 GMTModeling & metricsmetricstutorialmodelinghttps://datahashi.com/blog/2026-09-16-thin-layer-thick-warehouse/https://datahashi.com/blog/2026-09-16-thin-layer-thick-warehouse/What belongs in your warehouse versus your semantic layer, and why copying logic into both creates drift.Wed, 16 Sep 2026 00:00:00 GMTStrategy & comparisonssemantic-layerwarehousehttps://datahashi.com/blog/2026-09-15-metric-contracts-same-number-test/https://datahashi.com/blog/2026-09-15-metric-contracts-same-number-test/A metric contract is a shared definition you can test. If two consumers disagree on the number, the contract failed.Tue, 15 Sep 2026 00:00:00 GMTGovernance & trustgovernancemetricshttps://datahashi.com/blog/2026-09-14-the-engine-never-opens-your-warehouse/https://datahashi.com/blog/2026-09-14-the-engine-never-opens-your-warehouse/Why datahashi stopped holding warehouse passwords — compile and execute are different jobs, and agents make the wrong trust boundary expensive.Mon, 14 Sep 2026 00:00:00 GMTGovernance & trustarchitecturesecurityquery-runnerhttps://datahashi.com/blog/2026-09-14-why-llms-invent-joins/https://datahashi.com/blog/2026-09-14-why-llms-invent-joins/Explain why models invent joins from schemas and how governed metrics remove join invention from the agent loop.Mon, 14 Sep 2026 00:00:00 GMTAgents & MCPagentsllmsemantic-layerhttps://datahashi.com/blog/2026-09-13-one-request-shape-every-consumer/https://datahashi.com/blog/2026-09-13-one-request-shape-every-consumer/Dashboards, APIs, and agents should ask for metrics the same way. One structured request shape keeps definitions stable and tooling simple.Sun, 13 Sep 2026 00:00:00 GMTQueryingsemantic-layerapihttps://datahashi.com/blog/2026-09-12-semantic-layer-vs-dbt-vs-cubes/https://datahashi.com/blog/2026-09-12-semantic-layer-vs-dbt-vs-cubes/Compare dbt models, BI cubes, and a serving semantic layer so you know what each owns and why agents need a request-shaped contract on top.Sat, 12 Sep 2026 00:00:00 GMTStrategy & comparisonssemantic-layerdbtbihttps://datahashi.com/blog/2026-09-11-what-is-a-semantic-layer-for-agents/https://datahashi.com/blog/2026-09-11-what-is-a-semantic-layer-for-agents/A semantic layer gives agents stable metric definitions, governed access, and one query shape so AI systems stop inventing their own numbers.Fri, 11 Sep 2026 00:00:00 GMTAgents & MCPsemantic-layeragentshttps://datahashi.com/blog/2026-09-10-bi-semantic-to-agent-ready/https://datahashi.com/blog/2026-09-10-bi-semantic-to-agent-ready/How to move from a BI-only semantic layer to agent-ready serving without rewriting every dashboard.Thu, 10 Sep 2026 00:00:00 GMTStrategy & comparisonsbiagentssemantic-layerhttps://datahashi.com/blog/2026-09-09-querying-via-rest/https://datahashi.com/blog/2026-09-09-querying-via-rest/Call the semantic layer over REST with a metrics × dimensions × filters payload and read certified results - conceptual request and response shapes.Wed, 09 Sep 2026 00:00:00 GMTQueryingapiqueryingresthttps://datahashi.com/blog/2026-09-08-multi-tool-mcp-one-revenue/https://datahashi.com/blog/2026-09-08-multi-tool-mcp-one-revenue/A composite war story about conflicting revenue answers across MCP tools and how one certified metric restored trust.Tue, 08 Sep 2026 00:00:00 GMTAgents & MCPmcpagentswar-storyhttps://datahashi.com/blog/2026-09-07-postgres-mysql-starter-warehouse/https://datahashi.com/blog/2026-09-07-postgres-mysql-starter-warehouse/Use Postgres or MySQL as a starter warehouse for a thin semantic layer, then move to Snowflake or BigQuery without rewriting consumers.Mon, 07 Sep 2026 00:00:00 GMTWarehouses & setupwarehousetutorialpostgreshttps://datahashi.com/blog/2026-09-06-modeling-revenue-agent-safe/https://datahashi.com/blog/2026-09-06-modeling-revenue-agent-safe/How to define revenue as a certified metric with clear grain, filters, and guardrails so agents and humans get the same number.Sun, 06 Sep 2026 00:00:00 GMTModeling & metricsmetricsmodelingagentshttps://datahashi.com/blog/2026-09-05-when-you-dont-need-a-semantic-layer/https://datahashi.com/blog/2026-09-05-when-you-dont-need-a-semantic-layer/Honest signals for postponing a semantic layer versus when agent and multi-consumer chaos makes one urgent.Sat, 05 Sep 2026 00:00:00 GMTStrategy & comparisonssemantic-layerstrategyhttps://datahashi.com/blog/2026-09-04-audit-trails-agent-answers/https://datahashi.com/blog/2026-09-04-audit-trails-agent-answers/Implement audit trails for agent answers tied to metric requests so compliance can replay who asked what.Fri, 04 Sep 2026 00:00:00 GMTGovernance & trustgovernanceagentsaudithttps://datahashi.com/blog/2026-09-03-internal-data-agent-certified-metrics/https://datahashi.com/blog/2026-09-03-internal-data-agent-certified-metrics/Build an internal data agent that only answers from certified metrics via MCP, with a clear eval for same-number trust.Thu, 03 Sep 2026 00:00:00 GMTAgents & MCPmcpagentstutorialhttps://datahashi.com/blog/2026-09-02-vs-warehouse-native-semantics/https://datahashi.com/blog/2026-09-02-vs-warehouse-native-semantics/Contrast vendor-native semantic features with a portable serving layer that works across warehouses and agent tooling.Wed, 02 Sep 2026 00:00:00 GMTStrategy & comparisonssemantic-layerwarehousehttps://datahashi.com/blog/2026-09-01-claude-mcp-certified-metrics/https://datahashi.com/blog/2026-09-01-claude-mcp-certified-metrics/Configure Claude with MCP tools that only return certified metrics, keeping free-form SQL out of the loop.Tue, 01 Sep 2026 00:00:00 GMTAgents & MCPmcpagentsclaudehttps://datahashi.com/blog/2026-02-18-metering-every-query/https://datahashi.com/blog/2026-02-18-metering-every-query/datahashi records what each query scanned, so "what did this tenant cost" is one endpoint away — not a data-warehouse billing archaeology project.Wed, 18 Feb 2026 00:00:00 GMTGovernance & trustgovernancemeteringhttps://datahashi.com/blog/2026-01-15-introducing-datahashi/https://datahashi.com/blog/2026-01-15-introducing-datahashi/Same metric for humans and AI agents — without hallucinated SQL. A governed semantic layer you author as YAML, publish as an immutable artifact, and query over REST or MCP.Thu, 15 Jan 2026 00:00:00 GMTStrategy & comparisonsannouncementsemantic-layeragentsmcp