Populations as data
Treat the Population identity and its complete membership as data you can store, address and read back.
Developer PreviewEvolving data,
Evolving data,
governed state.
AxolDB is a general-purpose evolutionary population database management system. It keeps persistent facts, population transitions and lineage explicit—without confusing generated candidates with committed data.
Concept
How AxolDB models data.
Populations are identified collections; Generations preserve exact membership states; continuations remain explicit.
History you can inspect
Retain earlier Generations and compare saved states without replacing them with a single latest snapshot.
Alternative continuations
Resume different scenarios from one saved state while keeping their source reference explicit.
The Mesa links are application-level manifest provenance, not native Fork lineage.
Capabilities
What you can do with it.
Use explicit mappings for external tools, or run the bounded built-in Evolution operations demonstrated by the local qualification bundle.
Connect existing tools
Map application state deliberately, write it through public AxolDB operations and reload the exact records later.
This is explicit integration—not automatic import of arbitrary objects.
Native mutation and crossover
Run AxolDB's built-in mutation and binary midpoint crossover operators over candidate instances.
Verified with a local linux-x64 qualification bundle that is not a public release.
Trace parentage
Inspect ordered parent candidate instances, operator identity and the producing record for a derived candidate.
Compare analysis variants
Keep one source, exact rule versions and both result sets together while quanteda performs the text analysis.
The small GOV.UK benefit-guide corpus and dictionaries are illustrative, not a validated sociological instrument.
Verified examples
Concrete runs, with the boundary visible.
Each result below is tied to its retained run report. These are bounded demonstrations, not performance, scale or production-readiness claims.
DEAP checkpoint restore
The documented restored generation-50 path matches the uninterrupted path in ordered candidate bits, fitness, membership hashes and final metrics.
Mesa state restore
The documented run restores stable agent identities, groups, positions, both RNG states and activation order before continuing.
Its provenance is application data, not native Fork lineage.
Native Evolution
Mutation and binary crossover ran through one linux-x64 package; source state, result membership and lineage remained available after restart.
The package is a local qualification artifact, not a public Developer Preview download.
Quanteda rule comparison
The narrow dictionary labeled 16/20 sentence units and the expanded dictionary 18/20; two exact units changed and recomputation from AxolDB-loaded data matched.
Quanteda performs the text analysis; AxolDB stores and links source, rules and results. The local qualification bundle is not a public release.
See official downloads for current package availability and release statusfor exact identities and support boundaries. Native Evolution and Quanteda reports retain the qualification bundles they actually used.
Documentation preview
Understand the boundaries before you run it.
Agent SkillsInstall the separate MIT Architect and Operator packages; retained acceptance and scope.Release statusOfficial packages, download requirements and platform evidence boundaries.Core conceptsPopulations, Generations, Genotypes, Transactions, Forks and lineage.Install and runWindows ZIP and Linux tar.gz setup, requirements and integrity checks.