Show HN: Solvi 1.0 – an LLM proposes, checks decide, every decision replays

Solvi 1.0, a decision-system library from the solvi-ai project, launched with a design in which an LLM or classifier proposes answers while deterministic checks decide, and every decision carries a hash-chained trace that can be re-executed to confirm the answer or pinpoint an altered step. The library ships as pip extras including solvi[model] for ModernBERT field extractors, solvi[onnx] for CPU inference without torch, solvi[serve] for HTTP and MCP serving, and solvi[duckdb] for stored decisions, requiring Python 3.11+. On four benchmark tasks (Banking77, Abt-Buy, CUAD, RAGTruth) the solvi.build defaults matched or beat hand-written setups with every promise kept on eval, in a quarter to half of the code, and the slow path was rarely invoked.

Decision systems you can check. You describe what to compute with plain Python functions and checks; a model — an LLM, your own classifier or a small local one — proposes where judgement is needed; solvi's checks decide. Every answer comes with a confidence, a reason you can check the rule's inputs, a formula over computed facts, or a quote with character offsets in the source and a hash-chained trace that can be re-executed later to confirm the answer or pinpoint the step that was altered. When something cannot be computed or checked, solvi abstains or hands the case to a person instead of guessing; a failed hard check always overrides any model's confidence. solvi 1.0 has two levels. Ready systems you configure: solvi.build decisions from labelled examples, with a promise on the errors , solvi.Agent acting in an environment , solvi.Guard an agent's tool calls and solvi.Knowledge what they know, and from whom . Building blocks in solvi.core , which the ready systems are made of, each replaceable by a part of your own. What works but has no measured gain yet is kept apart in solvi.experimental . pip install solvi core: rules, checks, learned answer heads, build / Agent / Guard numpy, scipy, pydantic pip install "solvi model " + torch, transformers: ModernBERT field extractors for documents and the decider pip install "solvi onnx " + onnxruntime, tokenizers: the decider solvi.models.decider on CPU without torch pip install "solvi serve " + fastapi, uvicorn: solvi serve app.py:system — the questions over HTTP also --mcp pip install "solvi mcp " + the official MCP SDK for solvi serve --mcp without it, a built-in stdio server pip install "solvi duckdb " + duckdb: stored decisions in a DuckDB file solvi.core.store.DuckDBStorage ; postgres for PostgreSQL pip install "solvi lora " + torch, transformers, peft: solvi.experimental.lora.adapt lora, a LoRA adapter per question experimental Requires Python 3.11+ tested on 3.11–3.14 . Coming from 0.9: see below coming-from-09 . Give the question, labelled examples and the promise — and a slow path if you have one. solvi.build fits System 1, calibrates its guarantee and who answers what it hands over on examples it did not see, stores every decision, and says what it chose: python import solvi from solvi.models import llm s = solvi.build question, examples, catalog=cat, max risk=0.02, examples: state, correct answer slow=llm URL, "openai/gpt-oss-120b" , price= 0.037, 0.17 , storage="decisions.jsonl" res = s.ask state res.answer, res.by "s1", "s2" or "human" , res.reasons, res.cost print s.explain System 1, its signal and promise, who answers each slice, what is not covered On four tasks of the stand Banking77, Abt-Buy, CUAD, RAGTruth the defaults matched or beat the hand-written setups with every promise kept on eval, in a quarter to half of the code — closer to the promised level than the hand-written ones, and the slow path was rarely given anything guide https://github.com/solvi-ai/solvi/blob/main/docs/guide.md quick-start-solvibuild , examples/24 one entry point.py https://github.com/solvi-ai/solvi/blob/main/examples/24 one entry point.py . km = solvi.Knowledge "knowledge.jsonl", vocabulary={"here": lambda s, a: s "here" , "wood": lambda s, a: s "wood" } km.goal "table", done=lambda s: s "table" an agenda goal: done is code agent = solvi.Agent env, knowledge=km env: reset seed , actions state , step action → Outcome agent.run seed=7, steps=150 explore; the same world again needs fewer slow decisions agent.replay every decision re-checked from what it was given solvi.Knowledge keeps what a system learned and from whom: facts with their sources never the system's own answers , goals and gates, an action model learned from outcomes; a retraction takes back everything derived from the item. solvi.Agent acts on it: System 1 where the knowledge predicts the action works, a search where it does not, gates and predicted refusals as hard checks — protection by default, justified risk RiskBudget as an option. In the toy crafting world of examples/25 environment agent.py https://github.com/solvi-ai/solvi/blob/main/examples/25 environment agent.py the first run takes 61 steps, all decided by System 2, and the same world met again 12 steps, 11 of them by System 1. Growth was shown in environments met again, not on decision streams or a support agent with tools; justified risk lowers the cost of protection without promising parity docs/agent.md https://github.com/solvi-ai/solvi/blob/main/docs/agent.md . solvi.build ..., knowledge=km and solvi.Guard ..., knowledge=km read the same knowledge. The agent proposes {"name": tool, "arguments": {...}} ; solvi.Guard checks it — the tool is in the catalog, the arguments validate against its types, the values that must come from the conversation are quoted there and not only from a tool output that says "ignore previous instructions" , your policies limits, roles, allow-lists are ordinary hard checks, and an optional decider asks "did the user ask for this?" under act guard and perturb — then allows it solvi runs the function , denies it with the reasons, or escalates it to a person. Every decision is a stored, replayable trace; any framework's tool calls go through guard.check / guard.call guide https://github.com/solvi-ai/solvi/blob/main/docs/guide.md guarding-an-agents-tool-calls , examples/19 agent guard.py https://github.com/solvi-ai/solvi/blob/main/examples/19 agent guard.py . On τ-bench retail no call it let through was refused by the environment, at a cost in solved tasks the τ-bench row of the task table nine-public-tasks . Types declare questions; a model proposes; checks decide. The fields of a pydantic model are the questions, their types the kinds one option, several, an ordered score, yes/no ; a decider answers them about a text or a JSON / pydantic state with probabilities, a calibrated confidence and act / escalate, and hard checks, constraints and rules still decide. The decider is whichever model you have: - An LLM — solvi.models.llm base url, model, api key=... : any OpenAI-compatible chat-completions server OpenAI, OpenRouter, vLLM, llama.cpp, Ollama . Nothing to install beyond the core; its JSON replies are validated, and an invalid one escalates, never a guess. - A decision service — solvi.models.systemone url, model : anything that speaks the System One API. - A local checkpoint, for offline or cheap cases — solvi.models.decider "solvi-base" solvi onnx , no GPU; solvi models pull solvi-ai/solvi-base downloads it . solvi-base is a 150M ModernBERT-base cross-encoder distilled from solvi-large: about 50 ms per question on a CPU ONNX fp16, 4 threads . Its card is honest about where it stands: 54.5% zero-shot on typed questions over JSON states the same as solvi-large , 56.3% on Fast Decisions dev — not better than earlier small models there — and 0.602 on the jabr classifier benchmark, where Jev reaches 0.966. A preview: fit it on 30–60 labelled examples of your task and calibrate its escalation on your own stream before relying on it. python import os from typing import Literal from pydantic import BaseModel, Field from solvi import Catalog, Scale, System from solvi.models import llm class Triage BaseModel : team: Literal "billing", "technical", "shipping" = Field description="Which team should handle this ticket?" urgency: Scale Literal "low", "medium", "high", "critical" = Field description="How urgent is it?" angry: bool = Field description="Is the customer angry?" topics: list Literal "refund", "delay", "bug" = Field description="What does the ticket mention?" model = llm "https://api.openai.com/v1", "gpt-4o-mini", api key=os.environ "OPENAI API KEY" offline: model = solvi.models.decider "solvi-base" — the same questions, the same checks cat = Catalog questions = model.questions cat, Triage, text fact="ticket", min confidence=0.6 @cat.check hard=True, then={"urgency": "critical"} a legal threat is critical, whatever the model says def no legal threat ticket - bool: return "lawyer" not in str ticket .lower questions 1 .requires.append "no legal threat" res = System cat, questions .ask {"ticket": {"subject": "Charged twice", "body": "Refund my double payment ", "customer": {"tier": "pro"}}} print {q: r.answer, r.status for q, r in res.results.items } print res.audit "team" the probabilities, the model's fingerprint, what escalated and why A state is read as key paths customer.tier: pro ; an unsure or escalated answer abstains with the reason system.stats "model escalated" , "low confidence" . examples/15 typed decisions.py https://github.com/solvi-ai/solvi/blob/main/examples/15 typed decisions.py runs the whole story with a stand-in model, without network. Whatever the model, read what it was measured on before relying on it, and calibrate it on your own labelled stream part.act guard , under Building blocks : checks, constraints and escalation are what make the answers safe to act on, not the model alone. The checkpoint contract of a local decider is in docs/decide format.md https://github.com/solvi-ai/solvi/blob/main/docs/decide format.md ; a local checkpoint answers one question per forward pass with the published checkpoints, several with DecideModel.load ..., multi question=True . Every answer is a value and a confidence, and the types also declare answer primitives: Maybe T "not stated" — solvi.Unknown — is a real answer, unlike an abstention , Span float an exact piece of the text, parsed , Rank ... the top k, in order , Estimate 0, 7, 14 a number with an interval , and evidence quotes on any answer Claim value, evidence= ... , Question require evidence=True — each checked in the text, from rules or a model guide https://github.com/solvi-ai/solvi/blob/main/docs/guide.md answer-primitives-not-stated-evidence-spans-rankings-estimates , examples/16 primitives.py https://github.com/solvi-ai/solvi/blob/main/examples/16 primitives.py . - solvi playground https://huggingface.co/spaces/solvi-ai/playground : write a decision task in Python and run it, watch the strategist's plan, tamper with a trace and see the replay catch it, learn rules from examples. Its "New in 1.0" tab: solvi.build from labelled examples with explain , res.checks and then= as a function of facts, quotes matched on a normalized view, the compact journal, the two levels. - solvi documents https://huggingface.co/spaces/solvi-ai/documents : cited, typed answers from contracts, invoices, receipts, leases and more — the ModernBERT extractor ONNX and the decisions both run in your browser; add a field by describing it. - solvi arcade https://huggingface.co/spaces/solvi-ai/arcade : game agents that explain every move — tic-tac-toe, maze, minesweeper, 20 questions, Mafia detective, a bot arena, and "hack the trace". Its "Agent and knowledge 1.0 " tab: solvi.Agent in a small world — run 1 explores System 2 , run 2 in the same world walks what it learned System 1 , the knowledge report with a retraction, and protection vs justified risk RiskBudget behind a gate that holds. - solvi realms https://huggingface.co/spaces/solvi-ai/realms : an endless strategy game whose factions are solvi systems — tested for 100 000 turns: flat decision time ~0.3–0.6 ms , bounded memory and state, every sampled trace replay OK. - All run entirely in your browser Pyodide : no server, no GPU, nothing you type leaves the page. Each Space installs the solvi release it pins 1.0.0 . - Models: solvi-ai/solvi-large https://huggingface.co/solvi-ai/solvi-large typed decisions, 396M, preview , solvi-ai/solvi-base https://huggingface.co/solvi-ai/solvi-base the same answers on a CPU / in ONNX, 150M, preview , solvi-ai/extract-base https://huggingface.co/solvi-ai/extract-base fields by description and solvi-ai/extract-receipts https://huggingface.co/solvi-ai/extract-receipts . Each model card states what the model was measured on, how well it does, and its limits; all models are listed at huggingface.co/solvi-ai https://huggingface.co/solvi-ai . The ready systems are assembled from these; you can use them directly, and put a part of your own at every extension point a store, a slow path, a strategist, a head, a decider, an action model, ... . docs/building blocks.md https://github.com/solvi-ai/solvi/blob/main/docs/building blocks.md has one reference per protocol, with a conformance check for each. python from datetime import date from solvi import Answer, Catalog, Question, System cat = Catalog @cat.fn argument names = facts it reads; function name = fact it sets def days requested start, end : return end - start .days + 1 @cat.fn def remaining after balance, days requested : return balance - days requested @cat.check hard=True, then={"approve": "reject"} if this check is False, "approve" is forced to "reject" def enough balance remaining after : return remaining after = 0 @cat.check def enough notice start, today, days requested : return days requested < 5 or start - today .days = 14 @cat.rule "approve" def approve enough notice : return "approve" if enough notice else "needs manager" system = System cat, Question "approve", "Approve the leave?", Answer.choice "approve", "needs manager", "reject" , requires= "enough balance" res = system.ask {"start": date 2026, 10, 19 , "end": date 2026, 10, 23 , "today": date 2026, 9, 25 , "balance": 14} print res "approve" .answer, res "approve" .confidence, res "approve" .why print res.state text print res.trace.replay system Output: approve 1.0 enough notice = True days requested = 5 remaining after = 9 enough balance = True enough notice = True {'ok': True, 'steps': 5, 'mismatches': , 'models': , 'answers': 'same', 'catalog': 'same'} With "balance": 3 the hard check fails and the answer is reject with status == "forced" , whatever the rule says. solvi.show.show res, cat prints answers, the planned flow, the computed state and the replay result in one go. - Catalog. @cat.fn computation , @cat.check bool , @cat.extract value from text, returned as a Quote with offsets and @cat.rule question answer rule . A part's contract is its signature: argument names are the facts it reads, the function name is the fact it sets. Type hints are optional and become the facts' types def risk score risk points: dict str, float - float : producer and consumer types are checked when a part is registered, values are validated / coerced with pydantic at run time, and a value that fails is rejected like an ungrounded quote safeguard type rejected . Untyped parts cost nothing. - Questions. Question name, text, Answer.yes no | Answer.choice ... , requires= ... . Questions without a rule get a small answer head trained from labeled examples system.fit or a readable learned rule list system.learn rule . - Strategist. For each question it walks backwards from the rule's arguments or the learned features through the catalog signatures to the keys of init state , adds the question's required parts requires and every check that touches a computed fact. Everything else in the catalog is not executed; the flow records why each part was taken or skipped. - Execution. Each part runs once, even if several questions need it. Hard checks and their inputs run first; when one fails, the steps only the settled questions needed are skipped res.trace.skipped . With System ..., workers=8 independent steps run in parallel threads as soon as their inputs are ready. Results go into the computed state res.values , res.state text with their provenance; extracted values keep their quote. Each step record is hashed and chained to the previous one in flow order, so the trace does not depend on scheduling. - Answers and trace. res q .answer / .confidence / .why / .status ok , forced , abstain , plus res.trace.replay system , which recomputes every step from recorded inputs and reports mismatches, broken hash links and quotes outside the text — and, given the System, a stored answer that is not the one the trace gives. Grounded decisions. Fuzzy proposes, deterministic decides, everything is in the trace. Each fact and answer records its provenance — given , computed , quoted , decided a model's choice among options, with probabilities or learned — and model-backed steps record the model's id and fingerprint, so a replay can tell when the model changed since a decision. A model's quote that is not literally the text at its offsets, or a choice outside its options, is rejected and counted system.stats ; a fallback producer runs or the question abstains. print res.audit shows what each answer rests on, which safeguards fired, and how much of its support is deterministic. A decision without models and one with models are the same system examples/12 grounded audit.py https://github.com/solvi-ai/solvi/blob/main/examples/12 grounded audit.py . - A guaranteed risk. part.act guard examples, max risk=0.10 calibrates on a few hundred labelled examples of your stream so that P answered alone and wrong ≤ 10% for inputs like them conformal risk control — a share of all inputs, not the error among the answers given alone calibrate for max error=0.10, method="ltt" bounds that ; the audit shows the promise behind every answer, or says there is none. part.conformal examples gives the person who takes an escalation a short list of candidates. Near ties escalate min margin= , and the answer does not depend on the order the options are listed in sorted by default . - Several models. Any decider — an LLM, a System One service Jev, Kev, Von, Laya-serve, … , a local checkpoint — combines with the others: Cascade , Vote and Route solvi.core.deciders.combine — the next model only when one escalates, an answer only when models of different families agree, or a model picked by code — under one guarantee examples/20 vote across families.py https://github.com/solvi-ai/solvi/blob/main/examples/20 vote across families.py shows a vote with stand-in servers . - Serving and operations. solvi serve module:system exposes the questions over HTTP OpenAPI from the same types , MCP and the System One API; await system.aask ... runs async parts concurrently with timeouts. TraceStorage keeps decisions with a hash chain across them; solvi diff shows which stored decisions a rule or model change would flip; solvi test , solvi check and the honesty suite solvi honesty belong in CI; res.report format="html" and solvi report decisions.db --html out.html give an auditor one page per decision or per period; store.signature — 64 bytes kept next to the chain's head — later names the one stored record that was edited and restores its hash. - Text in. system.ask text "please refund order A-10457, 1.5 million RUB, paid on 12 September 2026", textin=TextIn system, decider, patterns={"order id": r"A-\d+"} : the decider picks which question the message asks or escalates when unsure , each input field is read with a quote found by a deterministic cue finder — chosen over the checkpoint's span pointer by measurement https://github.com/solvi-ai/solvi/blob/main/docs/guide.md text-in-from-a-message-to-a-question and a deterministic parser numbers, dates, enums, yes / no; a date without a year is not guessed , missing required fields are listed for a clarifying question, and the trace says those values were read by a model, not given. solvi serve answers texts at POST /ask text and as the MCP tool ask text . - Long documents. decider.decision ..., long="retrieve" : a contract longer than the decider reads is split into sections, BM25 picks the few that bear on the question, the decider reads only those, and quotes point into the whole document; the trace lists the sections read. long="full" reads a text whole up to the length a checkpoint trained on long inputs declares max len long , and retrieves within that length beyond it. - Corrections. fit heads refit on all kept examples as corrections accumulate; solvi.core.knowledge.memory.attach part escalates an answer when similar corrected cases say another one guide https://github.com/solvi-ai/solvi/blob/main/docs/guide.md a-memory-of-corrections-solvicoreknowledgememory . The audit, show and the safeguard report render in Russian with System ..., lang="ru" . A System under a guarantee is fast, cheap and knows when it is unsure: that is System 1. An LLM, a re-ask loop or a search is slow and costs money per input: System 2. solvi.build and solvi.Agent put them in one system; underneath: - Who answers solvi.core.dispatch . System 1 is asked first; when its own signals say its answer cannot be given alone — below its guarantee, unlike the calibration examples, an abstention, a broken constraint — the slow path answers, and when that cannot either, or the budget is spent, a person gets the input with both candidates and the reasons. Dispatcher.calibrate examples, max risk=... chooses, on each slice of what System 1 hands over, whether System 1's own guess, the slow path, their agreement or a person answers, under one promise. Every decision is one stored record with its cost and replays without calling a model guide https://github.com/solvi-ai/solvi/blob/main/docs/guide.md who-answers-system-1-the-slow-path-or-a-person-solvicoredispatch . - What happened System.report , solvi report decisions.db --overview . From the store alone: who answered and how often each handed over, the time, calls, tokens and dollars, the promise in force next to the error the stored labels show, and drift guide https://github.com/solvi-ai/solvi/blob/main/docs/guide.md the-system-report-systemreport-solvi-report---overview . - A showcase. A player walks the world map of Pokémon Red recorded from a real playthrough: place names and exits, no ROM through the first fifteen goals twice; System 1 is two rules over remembered routes, System 2 a search over the world map it writes as it goes, and after each goal what System 2 found becomes System 1's routes. The first run takes 147 slow decisions of 183, the second 2 of 62, the fewest moves possible; every decision replays examples/23 pokemon world map.py https://github.com/solvi-ai/solvi/blob/main/examples/23 pokemon world map.py , the replay viewer in spaces/pokemon https://github.com/solvi-ai/solvi/blob/main/spaces/pokemon . The default strategist needs the inputs of every producer of a fact. solvi.core.plan.cost.CostStrategist plans around producers whose inputs are never given, and with producers="equivalent" picks the cheapest verified plan by declared cost= an exact 0/1 program; hard checks that govern a question always stay in the plan . No model is involved; the plan is one hashed record in the trace and replay re-verifies it. python from solvi.core.plan.cost import CostStrategist system = System cat, questions, strategist=CostStrategist producers="equivalent" See docs/strategist.md https://github.com/solvi-ai/solvi/blob/main/docs/strategist.md and examples/17 cost strategist.py https://github.com/solvi-ai/solvi/blob/main/examples/17 cost strategist.py . With solvi model , fields are found by a fine-tuned ModernBERT extractor. The extracted value is always a span of the document, so every answer built on it can be cited. python from solvi.core.extract import LongSpanExtractor ex = LongSpanExtractor.load "solvi-ai/extract-base" a field is a description; long texts in 1024-token windows ex.fit text, "the total amount paid", start, end , ... , epochs=3 a few dozen labeled documents of your task ex.save "my-extractor" cat.extract ex.field "total", "the total amount paid" doc - Quote text, start, end, confidence , or no answer cat.extract ex.field "date", "the date of the purchase" @cat.fn def amount total : extracted values are strings; parse them in ordinary functions return float total.replace ",", "" LongSpanExtractor reads the whole text in overlapping windows and supports "no answer" with a per-field threshold ex.tune threshold name, held out . What the published solvi-ai/extract-base does without labels of your task, from its model card held-out fields and data sets, one seed : CORD receipt fields it was never trained on, 46.5% and 67.9%; five never-trained CUAD clause types, 73.1%; Kleister-NDA and SROIE, never seen, 2–95% by field addresses 2% . With per-field thresholds from 40 labeled contracts it reached 85.5% on CUAD; fine-tuned on 25–100 SROIE receipts, 86–89%. So describing a field is a start, not a finished extractor: label 25–100 documents and fine-tune. The receipt numbers below are a LongSpanExtractor too: solvi-ai/extract-receipts is extract-base fine-tuned on CORD and loads with LongSpanExtractor.load "solvi-ai/extract-receipts" its card says so; examples/07 receipts model.py https://github.com/solvi-ai/solvi/blob/main/examples/07 receipts model.py uses it . See docs/guide.md https://github.com/solvi-ai/solvi/blob/main/docs/guide.md extracting-fields-from-documents . Pieces that work and are tested but have not yet shown a measured gain. Importing one warns ExperimentalWarning , a decision that used one records it, and each one graduates or is removed by 1.2; solvi.experimental.STATUS and docs/experimental.md https://github.com/solvi-ai/solvi/blob/main/docs/experimental.md say what each is missing. - A policy text compiled into the catalog solvi.experimental.compile . An LLM writes plain functions, hard checks and rules from a policy; they are accepted without labelled examples only when every part cites its clauses, the code runs in a sandbox, two independent drafts agree on every generated input and tests derived from the text pass. A person settles what the drafts dispute, a changed text is recompiled with the stored decisions it moves listed. Agreement is not correctness: a misreading both drafts share is accepted guide https://github.com/solvi-ai/solvi/blob/main/docs/guide.md a-specification-compiled-into-the-catalog-solviexperimentalcompile . - Behind a coding agent's hooks solvi.experimental.hooks . solvi hook install puts solvi in front of Claude Code's edits and prompts: every Edit / Write is checked against a rules file forbidden patterns, required functions, Python calls read from the code; fuzzy questions for a model, which block only with a calibration and denied with the rule and the lines, sent to the user, or let through; a prompt gets the one project skill it needs, or nothing guide https://github.com/solvi-ai/solvi/blob/main/docs/guide.md solvi-behind-a-coding-agents-hooks , examples/22 coding agent hooks.py https://github.com/solvi-ai/solvi/blob/main/examples/22 coding agent hooks.py . - Verified charts solvi.experimental.charts : every number quoted from the text and checked unit, scale, a pie that adds up , as a deterministic SVG that replays to the same bytes examples/21 verified chart.py https://github.com/solvi-ai/solvi/blob/main/examples/21 verified chart.py . - Learning solvi.experimental.learning , solvi.experimental.lora , solvi.experimental.oncalib : a loop that promotes an update from trusted corrections only when it passes held-out, honesty and calibration gates, with rollback; a LoRA adapter per question on solvi-base solvi lora ; recalibration on the fly from outcomes, which does not keep the promise. - The MCP proxy solvi serve --guard catalog.py:guard --upstream CMD : the guard in front of an MCP server and counterfactuals solvi.experimental.counterfactual : the smallest input change that flips an answer . Every 0.9 import path still works in 1.0.x: it imports the same module and warns SolviDeprecationWarning with the path to use; 1.1 removes the old paths. solvi migrate PATH rewrites your code imports and dotted paths in strings, in .py and .md files ; solvi migrate PATH --check only reports. Stored decisions, calibration files and fingerprints do not change. The full table and what was removed: CHANGELOG.md https://github.com/solvi-ai/solvi/blob/main/CHANGELOG.md breaking-changes-and-migration . solvi init triage --with-model && cd triage a typed catalog, passing cases.json, README, CI workflow solvi test . && solvi check catalog.py:system regression cases and the catalog lint what CI runs solvi ask catalog.py:system example.json --audit one decision and what it rests on --json, --report html solvi models pull solvi-ai/solvi-base a local decider for offline use; solvi models lists, check measures solvi calibrate catalog.py:system route labels.csv --risk 0.1 act guard → route.calib.json, loaded by the catalog solvi hook install experimental Claude Code's edits checked against .claude/solvi-rules.toml solvi migrate src/ --check code still on the 0.9 import paths without --check: rewrite it Every command is in the guide https://github.com/solvi-ai/solvi/blob/main/docs/guide.md command-line . Every number in this README comes from a script in benchmarks/ https://github.com/solvi-ai/solvi/blob/main/benchmarks or from a published model card. Field extraction: the extract-receipts https://huggingface.co/solvi-ai/extract-receipts card reports 97.3% on typed questions over CORD receipts 100 test receipts with ECE 0.011 and 98.6% of questions answered at ≥ 99% precision, and the extract-base https://huggingface.co/solvi-ai/extract-base card its zero-shot and few-label numbers; the decision models' cards solvi-base https://huggingface.co/solvi-ai/solvi-base , solvi-large https://huggingface.co/solvi-ai/solvi-large theirs. On the gallery, the rules, quotes and checks are the answer: every answer is backed by a quote at stated offsets, a computed fact, or the system abstains. Details: docs/benchmarks.md https://github.com/solvi-ai/solvi/blob/main/docs/benchmarks.md . Real tasks on public data, each with a baseline that does not use solvi and one solution with solvi, chosen on dev and scored on a held-out split benchmarks/tasks/ https://github.com/solvi-ai/solvi/blob/main/benchmarks/tasks , solvi 0.8.0, gpt-oss-120b where an LLM is used : | Task | Baseline no solvi | solvi | |---|---|---| | CUAD contracts, 1,025 questions: accuracy; answered alone, wrong among them | 0.882; 100%, 11.8% | 0.899; 67.6%, 4.0% | | Banking77 stream, 20 unseen intents from request 1,000, promise ≤ 5% wrong: wrong after the shift | 22.5% broken | 0.7% kept , at 13.7% answered alone | | Abt-Buy, 1,916 product pairs: F1 | 0.872 LLM per pair | 0.933 code reads the offers; a head fitted on 5,743 labelled pairs | | NATURAL PLAN, right of 100: calendar / meetings / trips | 92 / 75 / 43 LLM plans | 95 / 100 / 98 solvi.core.slow.search , no LLM | | BIRD mini-dev, 150 questions: right; wrong among answered | 78; 48.0% | 73; 34.8% at 74.7% answered | | RAGTruth, 600 responses: F1 | 0.784 | 0.766 | | τ-bench retail, 30 tasks: solved; calls the environment refused | 18; 10 | 14; 0 | | NAB, 33 series: F1 | 0.391 | 0.361 | solvi does not make a model more accurate: on RAGTruth, BIRD, τ-bench and NAB it did not beat the baseline. What it added is the promise on what is answered alone, consistency across items, a search where a model guessed, and a stored, replayable record of every decision. German Credit rules and an audit of a rule change decides exactly as plain code does and adds the audit. Every number, its caveats and the cost of a run: benchmarks/tasks/README.md https://github.com/solvi-ai/solvi/blob/main/benchmarks/tasks/README.md . We gave the same inputs and written rules to solvi and to four LLMs: Grok 4.7, gpt-oss-120b, Qwen3-235B-2507 and DeepSeek-V3.2. The sets were the gallery, generated refund and 3-way-match cases built to trip models up, and Banking77 message routing. Strong reasoning LLMs followed short written rules as accurately as solvi. solvi was not more accurate there. What differs is everything around accuracy: | Refunds, 576 decisions | solvi | Grok 4.7 | gpt-oss-120b | Qwen3-235B-2507 | |---|---|---|---|---| | accuracy | 0.894 in the run, 1.000 with 0.7.0's claim reader not blind | 1.000 | 1.000 | 0.844 | | limit violations | 0 | 0 | 0 | 70 | | answers changed by reordering the options | 0% | 0% | 0% | 9.6% | | per decision | 0.6 ms, CPU | 2.0 s | 1.9 s | 1.1 s | | $ per 1,000 decisions | $0 | $1.30 | $0.12 | $0.03 | Inside solvi, no LLM broke a hard check, and wrong answers given without a person stayed within the promised 10%. On free text the LLMs win: 0.925-0.972 on bank messages against 0.675 for solvi-large. A hosted decision model, Jev by TypeSafe, read bank messages almost as well as the best LLM 0.964 but scored 0.819 on refunds with 53 limit violations; inside solvi it broke no hard check either. An open one that reasons first, Jeeves by PostHog, came closer on the rules 0.863 on refunds, 0.912 on 3-way match but still broke hard checks when asked directly. Full tables, caveats and a reproducible runner with every raw answer: docs/vs llm.md https://github.com/solvi-ai/solvi/blob/main/docs/vs llm.md , benchmarks/vs llm/ https://github.com/solvi-ai/solvi/blob/main/benchmarks/vs llm . One ask without a model benchmarks/ask speed.py https://github.com/solvi-ai/solvi/blob/main/benchmarks/ask speed.py : planning, running the parts and hashing the trace; Python 3.10, Intel i7-12700H shared with other jobs, medians of three runs of 15 passes : | case | steps | per ask | |---|---|---| | README quickstart leave request | 5 | 0.29 ms | | 15 gallery entries median entry; range | 4-23 | 0.88 ms 0.64-5.6 ms | | random catalog of 50 parts, 5 questions | 19 | 0.95 ms | | random catalog of 1 000 parts, 5 questions | 114 | 6.1 ms | | stream alert over an input of 3 712 floats the NAB task's catalog | 7 | 3.7 ms | A large input costs its hashing: every given value is written as canonical JSON and hashed once per ask about 0.25 µs per float . The slowest gallery entry checks its texts for instruction-like sentences perturb , which takes most of its time. Saving each response to a store adds its own cost the whole response is written as JSON . Strategist on random layered catalogs benchmarks/strategist scale.py https://github.com/solvi-ai/solvi/blob/main/benchmarks/strategist scale.py , one CPU core; solvi 0.7.1 with the changes since, Python 3.10, Intel i7-12700H, medians of two runs . The parts are trivial arithmetic: running all of them "run all" costs less than one ask, so the saving shows only when parts are slow services, models , as in the insurance desk below. | catalog parts | plan | plan + run + trace | parts run | share of catalog | run all, no plan | |---|---|---|---|---|---| | 100 | 0.14 ms | 1.1 ms | 31 | 31% | 0.03 ms | | 1 000 | 1.2 ms | 4.6 ms | 114 | 11% | 0.36 ms | | 10 000 | 11 ms | 25 ms | 232 | 2.3% | 4 ms | Insurance claim desk with six slow services of 100-300 ms examples/09 strategy at scale.py https://github.com/solvi-ai/solvi/blob/main/examples/09 strategy at scale.py : | questions asked | script computing everything | solvi, one by one | solvi, workers=8 | |---|---|---|---| | fast track? | 1 122 ms | 503 ms | 313 ms | | full decision + payout | 1 122 ms | 773 ms | 461 ms | | all five questions | 1 122 ms | 1 025 ms | 463 ms | | expired policy hard check settles it | 1 122 ms | 152 ms | 153 ms | A rule-only decision on a small catalog takes well under a millisecond benchmarks/ask speed.py https://github.com/solvi-ai/solvi/blob/main/benchmarks/ask speed.py ; on documents the extractor dominates. - The answer is a computation or a rule over a few values found in a document or a record: receipts, invoices, contracts, requests, orders. - You need to show why: auditors, compliance, or a human reviewing low-confidence cases. - Some rules are non-negotiable hard checks , and the rest can be learned from about 100 labeled examples. - Open-ended free-text questions or generated answers. solvi answers typed questions only: yes/no, choices, scores, multi-label, "not stated", exact spans of the text, rankings and number ranges. Around a model that writes a query, a plan, a JSON extraction it checks, compares and re-asks — solvi.core.slow.generate , solvi.core.slow.agree , solvi.core.slow.refine — but does not make the writing better. It searches for a plan only over a space you enumerate solvi.core.slow.search . - New fields with no labeled examples. Extracting a field from its description alone is not reliable yet: the published extract-base gets 2–95% by field on data sets it never saw see "Extract from documents" ; label 25–100 documents and fine-tune. - No labels at all. Plan on roughly 100 labeled documents field positions per task. - CPU-only deployment with a quantized extractor: dynamic int8 quantization changed half of extract-base's spans its model card , so none is provided; use fp32 or fp16 and measure the time per document on your hardware. gallery/ https://github.com/solvi-ai/solvi/blob/main/gallery — fifteen decision tasks across directions support triage, email routing, content guard, security alerts, AI-agent audit, release rollout, KYC/AML, clinical screening, credit with adverse-action reasons, procurement 3-way match, double-charge refunds, predictive maintenance , each with scenarios, a runner and a side-by-side against an answer-only model. Three are helpers for coding agents: a pre-edit rule check allow / block / escalate a file write , review triage quick review only when seven risk questions are a confident "no", with a stated bound on risky changes that slip through and a skill picker with an honest "none". | File | What it shows | |---|---| | examples/01 leave request.py https://github.com/solvi-ai/solvi/blob/main/examples/01 leave request.py | Leave request from a plain dict: rules, hard checks, parts the strategist skips | | examples/02 shop order.py https://github.com/solvi-ai/solvi/blob/main/examples/02 shop order.py | Shop order: two rule-based questions plus a "suspicious?" question learned from history with fit | | examples/03 invoices.py https://github.com/solvi-ai/solvi/blob/main/examples/03 invoices.py | Invoice approval from text: regex extractors with quotes, four questions, one learned | | examples/04 refunds.py https://github.com/solvi-ai/solvi/blob/main/examples/04 refunds.py | Refund e-mails: yes/no learned from examples under a hard "within 30 days" check | | examples/05 tic tac toe.py https://github.com/solvi-ai/solvi/blob/main/examples/05 tic tac toe.py | Tic-tac-toe agent: each move is an answer with its reason win, block, fork, ... ; a hard check rejects invalid boards | | examples/06 learned rules.py https://github.com/solvi-ai/solvi/blob/main/examples/06 learned rules.py | learn rule : route parcels to delivery zones from free-form addresses with a readable if-then list learned from labels | | examples/09 strategy at scale.py https://github.com/solvi-ai/solvi/blob/main/examples/09 strategy at scale.py | Insurance claim desk: the strategist generates a different plan per question set, hard checks first with early exit, slow services in parallel; timed | | examples/10 learn in milliseconds.py https://github.com/solvi-ai/solvi/blob/main/examples/10 learn in milliseconds.py | fit : a new question learned in milliseconds, then corrected one example at a time each correction ~0.2 ms, nothing retrained | | examples/11 answer types and constraints.py https://github.com/solvi-ai/solvi/blob/main/examples/11 answer types and constraints.py | Multi-label and ordinal answers tied by constraints between answers; contradictions in learned answers are repaired by joint decoding | | examples/12 grounded audit.py https://github.com/solvi-ai/solvi/blob/main/examples/12 grounded audit.py | One catalog with and without models: provenance, res.audit , a hallucinated quote caught by grounding, a decision outside its options, a model changed since the decision, lifetime safeguard stats | | examples/13 decide model.py https://github.com/solvi-ai/solvi/blob/main/examples/13 decide model.py | Support-email routing by a decider model as a catalog part: bias correction on unlabelled emails, 16 labelled examples, abstention, a constraint with a rule-based question, a hard check, the audit, teach , escalation for a target error rate, a JSON ticket the real model with SOLVI DECIDE MODEL , a stand-in otherwise | | examples/14 typed catalog.py https://github.com/solvi-ai/solvi/blob/main/examples/14 typed catalog.py | Typed facts: a pydantic request, type hints as fact types, answer types from the rules' return types Enum, Literal, bool , a mismatch caught at registration, rejected values → fallback / abstention, a response as JSON that loads back and replays | | examples/15 typed decisions.py https://github.com/solvi-ai/solvi/blob/main/examples/15 typed decisions.py | Typed decisions: a pydantic ticket, the questions as a pydantic model's fields choice, ordinal score, yes/no, multi-label , four answers one forward pass when the model shares passes , a hard check, a constraint and a rule over the model, an escalation in the audit and stats the real model with SOLVI DECIDE MODEL , a stand-in otherwise | | examples/16 primitives.py https://github.com/solvi-ai/solvi/blob/main/examples/16 primitives.py | Answer primitives: "not stated" vs abstain, spans parsed into numbers, evidence quotes checked in the text require evidence , a ranking with scores, an estimate with an interval — from rules and from a decider with the answer-primitives contract; confidence per kind, JSON round trip, replay | | examples/17 cost strategist.py https://github.com/solvi-ai/solvi/blob/main/examples/17 cost strategist.py | The code strategist: dead ends dropped, the cheapest verified plan by declared costs, the plan in the trace | | examples/18 several models.py https://github.com/solvi-ai/solvi/blob/main/examples/18 several models.py | Several models, one decision: a cascade small → large, a vote of two model families, a route by code — each under one act guard guarantee, with cost per question; every stage in the audit and the trace | | examples/19 agent guard.py https://github.com/solvi-ai/solvi/blob/main/examples/19 agent guard.py | An accounts-payable agent's tool calls through a Guard : grounded arguments, an invented IBAN denied, a budget escalation approved by a person, an instruction hidden in an invoice, an authorizer with act guard and perturb ; every decision stored and replayed a scripted agent, no API keys | | examples/20 vote across families.py https://github.com/solvi-ai/solvi/blob/main/examples/20 vote across families.py | A vote of two model families behind the System One API stand-in servers started in-process : each alone and the vote under one act guard guarantee; a sure mistake of one family escalates; the audit and the replay | | examples/21 verified chart.py https://github.com/solvi-ai/solvi/blob/main/examples/21 verified chart.py | A verified chart solvi.experimental.charts , experimental : every number quoted from the text and checked; a careless model's swapped digit, invented share and unquoted value dropped with reasons; a deterministic SVG that replays to identical bytes | | examples/22 coding agent hooks.py https://github.com/solvi-ai/solvi/blob/main/examples/22 coding agent hooks.py | A coding agent's session behind solvi hook : the hooks installed in a temporary project, a clean edit allowed, an edit that takes an employee id from the browser denied with the rule and the line, a migration with an empty downgrade and a comment that tries to talk past the rules denied, a skill picked for one prompt and none for another; the store verified and one decision audited and replayed | | examples/23 pokemon world map.py https://github.com/solvi-ai/solvi/blob/main/examples/23 pokemon world map.py | System 1 and System 2 on the world map of Pokémon Red recorded, no ROM : rules over remembered routes, a search over the player's world map when they are unsure or surprised, routes compiled after each goal; 147 → 2 slow decisions from the first run to the second; every decision stored, replayed and reported System.report | | examples/24 one entry point.py https://github.com/solvi-ai/solvi/blob/main/examples/24 one entry point.py | one entry point solvi.build : System 1 fitted from labelled examples, its guarantee and the slow path's slice calibrated on examples it did not see, every decision stored; explain prints the choices no model | | examples/25 environment agent.py https://github.com/solvi-ai/solvi/blob/main/examples/25 environment agent.py | An environment agent solvi.Agent with solvi.Knowledge on a toy crafting world: System 1 acts on what the action model and the skills predict, System 2 searches, the agenda's gates are hard checks; the same world met again takes 12 steps instead of 61 11 by System 1 , and protection vs RiskBudget at a bridge that breaks one time in three; every decision replays | | examples/07 receipts model.py https://github.com/solvi-ai/solvi/blob/main/examples/07 receipts model.py | Expense check on a scanned receipt: a receipts-tuned extractor cites each field, rules and a hard check decide needs solvi model | | examples/08 contracts by description.py https://github.com/solvi-ai/solvi/blob/main/examples/08 contracts by description.py | Contract review with fields defined only in words: the general extractor reads the whole contract, cites clauses or says "absent" needs solvi model | Run them from a clone: python examples/01 leave request.py . - docs/using.md https://github.com/solvi-ai/solvi/blob/main/docs/using.md : the ready systems — build , Agent , Guard , Knowledge and solvi.models . - docs/agent.md https://github.com/solvi-ai/solvi/blob/main/docs/agent.md : agents and knowledge, with what was and was not shown. - docs/building blocks.md https://github.com/solvi-ai/solvi/blob/main/docs/building blocks.md : the extension points of solvi.core . - docs/experimental.md https://github.com/solvi-ai/solvi/blob/main/docs/experimental.md : what is experimental and what each piece is missing. - docs/guide.md https://github.com/solvi-ai/solvi/blob/main/docs/guide.md : the reference guide, in the same three parts. - docs/best practices.md https://github.com/solvi-ai/solvi/blob/main/docs/best practices.md : what we learned while building on solvi, as advice. - docs/mission.md https://github.com/solvi-ai/solvi/blob/main/docs/mission.md : what solvi is for, what it is not, and its honest limits. - docs/decide format.md https://github.com/solvi-ai/solvi/blob/main/docs/decide format.md : the decider checkpoint contract for training your own . - docs/benchmarks.md https://github.com/solvi-ai/solvi/blob/main/docs/benchmarks.md : the scripts behind the numbers, what solvi adds to an ask, the dataset loaders. - docs/vs llm.md https://github.com/solvi-ai/solvi/blob/main/docs/vs llm.md : solvi vs asking an LLM Grok 4.7, gpt-oss-120b, Qwen3, DeepSeek , with raw answers. - ROADMAP.md https://github.com/solvi-ai/solvi/blob/main/ROADMAP.md , CHANGELOG.md https://github.com/solvi-ai/solvi/blob/main/CHANGELOG.md , CONTRIBUTING.md https://github.com/solvi-ai/solvi/blob/main/CONTRIBUTING.md , SECURITY.md https://github.com/solvi-ai/solvi/blob/main/SECURITY.md . - benchmarks/ https://github.com/solvi-ai/solvi/blob/main/benchmarks : the benchmark scripts that can be rerun, and the dataset loaders behind the extraction numbers SROIE, CORD, CUAD, Kleister-NDA — their scripts are not in the repository . - Tests: pytest . Apache-2.0. See LICENSE https://github.com/solvi-ai/solvi/blob/main/LICENSE . A paper is in preparation. Until then, please cite the repository: @software{solvi, title = {solvi: verifiable decision systems from catalogs of functions and checks}, author = {mxkuzn and solvi contributors}, year = {2026}, url = {https://github.com/solvi-ai/solvi} }