--- title: ANNForge: Approximate Nearest-Neighbor Vector Search | SpinForge url: https://spin-forge.com/forge15/annforge description: Approximate nearest-neighbor search over large vector collections, with recall checked against exact reference queries. --- FORGE15 / 03 / SEARCH & DATA [The full offering](https://spin-forge.com/platform) # ANNForge. Approximate nearest-neighbor search over large vector collections, with recall checked against exact reference queries. [Discuss ANNForge](https://spin-forge.com/contact?product=forge15%2Fannforge) Customer outcome [Operate compute and data workflows](https://spin-forge.com/platform/compute-data-operations) Available now CLI MCP Headless Choose your access mode THE PRODUCT, IN PRACTICE ## Find useful neighbors. Know what retrieval missed. A large vector collection creates a trade-off between retrieval cost and how many of the true nearest neighbors are found. ANNForge provides an approximate search path and a reference-based recall check. The useful output is not only a returned neighbor list: it is a result whose retrieval quality can be evaluated at the customer’s dataset, query and resource scale. BUILT FOR Vector-search, retrieval and data-platform teams evaluating a supported nearest-neighbor workload. CHALLENGES PRODUCT RESPONSE ## The problem you face. What ANNForge changes. 01 ### Exact search is too costly for the evaluation budget Scanning every candidate for every query may not suit the intended vector-search workload. ANNForge offers an approximate-neighbor path for the supported vector collection, giving the team another retrieval configuration to evaluate. 02 ### Fast search can omit relevant neighbors Query speed alone does not say how well an approximate result matches the true nearest set. The recall@k checking path compares returned neighbors with exact reference search on the evaluation queries. 03 ### A benchmark may not represent the customer’s data Recall, dimensionality, query count and index cost belong to the same deployment decision. Qualification uses the actual dataset characteristics and query workload to compare retrieval quality, runtime and memory together. WHAT YOU GET ## A defined capability. A useful output. The product scope below describes the capability and outputs to confirm for your release. Custom integration, operated delivery and additional engineering are separately agreed. 01 ### Approximate neighbor retrieval Search results for the supported vectors and query set. 02 ### Neighbor identities The returned approximate nearest-neighbor selections for downstream retrieval use. 03 ### Recall evaluation A comparison with exact reference neighbors on the qualification queries. 04 ### An integration-ready scope Defined vector dimensions, k, input size and resource requirements for evaluation. AVAILABLE NOW ## One product. Three ways to work. ANNForge is fully available with CLI, MCP and headless access. Use the interface that fits your terminal, your agents or your automated workflow. 01 / CLI ### For your terminal. Run the product from a terminal or script, alongside the tools your team already uses. [Get CLI access](https://spin-forge.com/contact?product=forge15%2Fannforge&interface=cli&brief=I+would+like+ANNForge+access+through+CLI.+Please+share+the+setup+documentation+and+configuration+for+the+selected+product.+We+will+confirm+the+environment%2C+license+scope+and+workflow+during+onboarding.) 02 / MCP ### For your agents. Connect MCP-compatible clients and agents to the product’s exposed capabilities. [Get MCP access](https://spin-forge.com/contact?product=forge15%2Fannforge&interface=mcp&brief=I+would+like+ANNForge+access+through+MCP.+Please+share+the+setup+documentation+and+configuration+for+the+selected+product.+We+will+confirm+the+environment%2C+license+scope+and+workflow+during+onboarding.) 03 / HEADLESS ### For your workflows. Operate without a graphical interface in unattended, scheduled or pipeline-driven work. [Set up headless access](https://spin-forge.com/contact?product=forge15%2Fannforge&interface=headless&brief=I+would+like+ANNForge+access+through+headless.+Please+share+the+setup+documentation+and+configuration+for+the+selected+product.+We+will+confirm+the+environment%2C+license+scope+and+workflow+during+onboarding.) Product-specific inputs, hardware requirements and result guarantees still apply. An access mode does not change your license or deployment scope. [Explore all product interfaces](https://spin-forge.com/developers?section=access-directory) EXAMPLE CUSTOMER WORKFLOW ## Compare retrieval configurations on a vector corpus A search team wants to test a different retrieval path without accepting a speed claim that omits quality. Illustrative use of the supported scope, not a customer case study or a measured result. 1. 01 Choose the supported corpus, queries and requested neighbor count. 2. 02 Run approximate retrieval and exact reference checks on the evaluation set. 3. 03 Compare recall, query cost and memory at the same workload scope. IN YOUR WORKFLOW ## Start with your task. Agree the useful outcome. Vector-search, retrieval and data-platform teams evaluating a supported nearest-neighbor workload. What you bring The vector corpus, dimensions, query set, k, acceptable recall and execution constraints. What a useful evaluation measures - Recall@k against exact reference queries - Query and preparation time within disclosed boundaries - Memory footprint for the chosen corpus Where it fits Use the selected product in a qualified customer environment, or agree a defined implementation with SpinForge. A hosted or managed service is quoted for its actual operating responsibilities; it is not automatically included with software access. QUESTIONS BEFORE YOU BUY ## A clear scope. No hidden assumptions. **How can we use ANNForge?**CLI, MCP and headless access are all available. Run from a terminal or script, connect an MCP-compatible client, or use the product without a graphical interface. Choose the mode during onboarding and receive its setup instructions. Runtime requirements and supported operations remain product-specific. **What is the important product boundary? **Approximate retrieval is not an exact-neighbor guarantee for every query. The supplied 0.821 recall on five million vectors belongs to its specific campaign settings; production comparisons need the dataset, dimensions and timing scope. **What comes with the selected license? **The agreed ANNForge release, its supported interfaces and outputs, the applicable checks and the support stated in your order. The plan defines users, workflow, environment and capacity. Source-code transfer, unrestricted redistribution and bespoke engineering are not automatically included. **Can SpinForge implement or operate it for us? **Request a scoped engineering or operated-service proposal. It identifies what our team delivers, where the workload runs, infrastructure costs, acceptance requirements and ongoing support. Your software subscription alone does not activate a hosted runtime. [Explore engineering](https://spin-forge.com/engineering) **What evidence is available for review? **P47 / Company-reported. The supplied campaign reports recall of 0.821 on five million vectors in 36.85 seconds. The dataset, dimensions, query count, k and timing boundary must accompany any comparison. These descriptions come from the supplied product and campaign material. Website checks are not an independent rerun of the engine. Qualification identifies the actual release, workload and available artifacts. [Read the evidence summary](https://spin-forge.com/results) SOFTWARE / OPERATED DELIVERY By proposal Customer-hosted access and a service operated by SpinForge are different scopes. Confirm the release, deployment and included support before activation. [Plans & delivery](https://spin-forge.com/pricing) PUT ANNFORGE TO WORK ## Tell us what needs to improve. Bring a representative workload and the result your team needs. We’ll define the product, deployment and engineering scope around that task. [Discuss this workflow](https://spin-forge.com/contact?product=forge15%2Fannforge&brief=Product%3A+ANNForge%0AOur+challenge%3A+Exact+search+is+too+costly+for+the+evaluation+budget%0AWorkload+%2F+baseline%3A+%0AEnvironment%3A+%0ARequired+output%3A+%0AAcceptance+measures%3A+Recall%40k+against+exact+reference+queries%3B+Query+and+preparation+time+within+disclosed+boundaries%3B+Memory+footprint+for+the+chosen+corpus%0ADelivery+preference+%28software+%2F+engineering+%2F+managed%29%3A+) **Related products, workflows and system connections + **RELATED CAPABILITIES ## Connect the right products. Distinct responsibilities. Explore each scope before choosing a combined deployment. [### GraphForge Graph-analysis software for PageRank, connected components and shortest paths, with independent result checks.](https://spin-forge.com/forge15/graphforge) [### sketchstream Streaming frequency and distinct-count estimation designed for large streams with a small memory footprint.](https://spin-forge.com/forge15/sketchstream) [### Search & streaming data Search vectors, summarize streams and compute graph results with checks suited to each output.](https://spin-forge.com/solutions/search-data) ONE PLATFORM / OR A DIRECT PURCHASE ## Checked Decisions with ANNForge. Activate this capability inside a qualified checked decisions workflow. Compatible methods and required shared checking are included once within that workflow; the product remains independently purchasable. Capacity, release, environment, result contract and operating responsibilities are specified before activation. Choosing an additional method is not permission to add an unrelated workload or unqualified deployment. [Explore Checked Decisions](https://spin-forge.com/platform/checked-decisions) [Choose workflow capacity](https://spin-forge.com/plans?configuration=checked-decisions&product=forge15/annforge) CONNECTED WORKFLOWS ## Use ANNForge in your workflow. The product remains independently available. These customer-specific solution paths use the relevant supported capabilities, with integration and delivery rights agreed separately. [05 ### Performance & release checks Make performance and quality evaluation a repeatable part of your release process. Compare an agreed workload before a change is accepted into production.](https://spin-forge.com/solutions/release-checks) [Explore all 17 solution workflows](https://spin-forge.com/workflows) FROM PRODUCT TO CUSTOMER WORKFLOW ## A useful outcome. A defined delivery. Commission the model, integration and acceptance work needed for your task. Existing products stay independently available. [03 / Scoped customer program ### Improve a computational workflow Commission a bounded performance improvement with source identity, numerical acceptance and the appropriate checking evidence.](https://spin-forge.com/programs/compute-improvement) [Explore decision programs](https://spin-forge.com/decision-systems) THE WHOLE-SYSTEM ROLE ## ANNForge in a wider workflow. Retrieve candidate reusable proofs, policies, configurations or diverse prior solutions. POTENTIAL SYSTEM CONTRIBUTION Similarity is never a validity proof; misses reduce reuse, not correctness, if fallback is sound. [S3 · Checked computational memory](https://spin-forge.com/system-synergies/checked-computational-memory) [S4 · Options, flexibility & recourse](https://spin-forge.com/system-synergies/options-and-recourse) [S6 · Checked improvement feedback](https://spin-forge.com/system-synergies/checked-improvement) [See this product in the map](https://spin-forge.com/system-synergies/product-map?q=ANNForge) [Define an integration evaluation](https://spin-forge.com/system-synergies/evaluation) The product is independently available through CLI, MCP and headless access. The joins above are integration candidates; compatibility, authority, data rights and end-to-end benefit must be established for the selected workflow. START WITH THIS PRODUCT Use the software directly or ask us to scope the implementation. [Software access](https://spin-forge.com/plans?product=forge15/annforge) [Have us do the work](https://spin-forge.com/contact?product=forge15/annforge&delivery=engineering)